FAQ: Common Questions and Common Failure Modes

Q. What does Mind-Upload actually do?

A. It is a site for building a Verification Commons that turns mind uploading and WBE into a verifiable research program. It fixes the data inputs, evaluation outputs, rules for success and failure, and operational procedures first.

Q. Can EEG read “thoughts”?

A. It is fair to say that some aspects can be read under constrained conditions, but it is not yet fair to say that free thought can be read as such. The reason is that the strongest non-invasive papers do not all solve the same problem. Tang et al. (2023) demonstrated subject-cooperative semantic reconstruction from within-subject fMRI. Défossez et al. (2023) showed 3 s speech-segment identification from non-invasive M/EEG with predictions dominated by lexical and contextual representations. d'Ascoli et al. (2025) showed known-word-onset decoding across 723 people and roughly five million words, but performance still depended strongly on task structure, with MEG and reading outperforming EEG and listening. Ye et al. (2025) then showed prompt-conditioned fMRI-to-LLM generation that beat a permuted-brain control, but still depended heavily on prompt and LLM scaffold.

Another correction is needed at the task-design layer itself. Rybár et al. (2024) showed that EEG semantic-category decoding can reach mean accuracies up to 71.3% during cue presentation while failing to decode the same categories during the later, cue-separated mental-task period. So a paper that mixes cue-period data with internally generated processing can overstate the communication ceiling before any decoder architecture is discussed. Horikawa (2025) then added a different non-invasive route again: whole-brain fMRI was linearly decoded into semantic features and iteratively optimized into captions of viewed or recalled videos. That is a viewed/recalled content captioning route, not the same object as continuous language reconstruction, known-onset word decoding, or free-running language production.

It is also not the case that scalp signals uniquely determine the internal state. In Unnwongse et al. (2023), which used intracranial stimulation for direct validation, the mean ESI localization error ranged from 10.3 to 26.0 mm depending on source depth and skull conductivity. In Hao et al. (2025), using simultaneous HD-EEG and SEEG, ictal ESI outperformed interictal ESI, but the figures were still 14.07 ± 4.62 mm versus 17.38 ± 4.16 mm, and accuracy depended strongly on source depth and spike power.

There is also an earlier limit than the inverse solver: not every neural pattern creates a usable scalp field. Ahlfors et al. (2010) showed that sensitivity already changes with source orientation, Ahlfors et al. (2010) showed that extended sources can cancel at the surface, Goldenholz et al. (2009) showed that anatomy and source extent can shift SNR by around 10 dB in mesial temporal examples, and Piastra et al. (2021) showed that omitting CSF overestimates EEG sensitivity. So a prettier map or more channels do not by themselves create observability that field formation never provided.

On this site, we therefore do not treat “externally validated ESI” as one checkbox. Intracranial-stimulation ground truth, simultaneous HD-EEG/SEEG, and postsurgical outcome answer different error questions, so the benchmark class has to be named before the claim ceiling is raised. The shortest route is EEG 101: what must now be stated rigorously about ESI and Datasets: the source-imaging validation ladder.

Another weak reading is to publish only the cleanest inverse map and silently ignore how much the answer moves across reasonable pipelines. Mahjoory et al. (2017) showed that inverse-method and software-package choice can materially change source localization and even more strongly change downstream connectivity estimates. Mikulan et al. (2020) then showed on a ground-truth intracranial-stimulation benchmark that many tested parameter combinations miss the session-wise optimum by a wide margin. So at Mind-Upload, one pretty ESI map without cross-solver / cross-parameter spread or an uncertainty display is still read as method-sensitive evidence, not as a stable source fact.

The Mind-Upload position is not to deny ambitious readout work. It is to separate the claim first into task-dependent decoding and internal-state identification, then make explicit the language prior, calibration, abstention conditions, and whether direct validation exists. If you skip those distinctions, you end up misreading “a string was produced” as if it were WBE-relevant state reconstruction.

Q. What is the minimum you should check in a brain-to-text demo?

A. At minimum, check the following eleven things.

• Measurement method: scalp EEG, MEG, fMRI, ECoG, or intracortical array. Representative high-performance speech neuroprostheses are invasive.
• Task: heard words, read words, speech articulation, recall, or free conversation. A constrained perceptual task is not the same as free thought.
• Cue regime: whether the paper decodes cue presentation, externally supported perception, or a cue-separated self-generated mental period. Cue-period performance is not the same thing as self-generated semantic BCI performance.
• Timing and segmentation: known word onset, fixed multi-second segment, fMRI TR window, prompted continuation, or free-running onset detection. Known onset is not the same thing as unconstrained decoding.
• Output family: fixed-bank segment retrieval, known-onset word decoding, prompt-conditioned generation, viewed/recalled content captioning, or streaming attempted-speech synthesis. These outputs do not share one uncertainty object.
• Priors and baselines: fixed vocabulary, beam search, external corpora, LLMs, prompt length, and how far LM-only, no-brain, permuted-brain, no-text-prompt, or shuffle baselines were used. Fluency does not automatically reflect brain signal alone.
• Validation: held-out conditions, counterfactual tests, adversarial controls, and whether failure cases are shown. Evaluation that stays too close to the training setup is not strong evidence.
• Subject route: within-subject, cross-subject, adapted-to-subject, or zero-shot to unseen people, together with any cooperation requirement or countermeasure test.
• Operational route: same-session throughput / expressivity, transfer-assisted initialization, fixed-decoder durability, and adaptive rescue answer different questions. Do not collapse them into one `speech BCI` score.
• Confidence handling: whether confidence is calibrated, and whether silence or abstention is returned at low confidence. A high-probability display alone is not safe interpretation.
• Long-term operation: not just within-session speed, but also tail latency, cross-day stability, and recalibration burden. A fast demo is not the same as a deployable loop.

One more front-door error had to be blocked here: papers that output fluent sentences can still be solving different problems at the task-design level. Rybár et al. (2024) showed that cue-period data can create apparently strong semantic BCI performance that does not survive into the separated mental-task window. Horikawa (2025) showed descriptive caption generation from viewed or recalled video content through semantic-feature decoding and iterative text optimization. Those are real advances, but they do not sit on one monotonic `brain-to-text` ladder with known-onset word decoding, prompt-conditioned language continuation, or invasive speech communication.

The front door was still too coarse when it let invasive language BCIs sound like one continuous success story. Willett et al. (2023) strengthened the same-session throughput route and also showed a bounded no-new-day-training slice (30% word error rate offline without new-day retraining), but the authors still said clinically viable multi-day adaptation remained unfinished. Littlejohn et al. (2025) strengthened streaming throughput / expressivity. Wairagkar et al. (2025) strengthened instantaneous voice synthesis with silence fallback, but even its session-1 decoder used same-day training. Singh et al. (2025) instead strengthened a cross-subject transfer / initialization route, and Karpowicz et al. (2025) plus Wilson et al. (2025) strengthened adaptive stabilization / unsupervised rescue. Those are different operational achievements, so general scalp EEG or ordinary non-invasive BCI cannot claim the same level without naming which route actually improved.

Q. If the score is high or cross-day stable, did we read the target neural variable?

A. Not necessarily. A score can stay high because the model is using shortcut routes rather than the intended neural variable. In EEG/BCI work, those routes can include eye position, facial/jaw/neck EMG, uninstructed movement, auditory feedback, subject / session fingerprint, and acquisition-distribution cues such as site, device, reference system, electrode layout, and protocol.

Musall et al. (2019) showed that trial-by-trial neural dynamics can be dominated by richly varied movements, Mostert et al. (2018) showed that visual-working-memory decode can retain an eye-movement confound, McFarland et al. (2005) showed that EMG can contribute to early BCI performance, Chaibub Neto et al. (2019) showed that repeated-measure record-wise splits can massively underestimate error through identity confounding, and Xu et al. (2020) showed that cross-dataset variability weakens EEG-decoding generalization.

At Mind-Upload, a result is not read as target-specific neural evidence unless the paper also fixes the target variable, nuisance-only baselines, slice-wise hold-out, and independence units for subject / session / acquisition distribution. The shortest follow-up is Verification: Specificity & Shortcut Card.

Q. If a foundation-model or leaderboard result is strong, did we solve general EEG decoding?

A. Not by default. On this site, a large EEG model or leaderboard result is first read as benchmark-conditioned transfer evidence, not as automatic proof that a general neural decoder already exists. The current literature already splits that evidence into at least four different objects. Jiang et al. (2024) already treated mismatched electrodes, unequal sample lengths, varied task designs, and low signal-to-noise ratio as core EEG-side barriers even while reporting LaBraM, and Lee et al. (2025) then found only marginal gains, about 0.5%, over conventional deep baselines despite much larger parameter counts. The newer setup-agnostic papers matter, but they matter in a narrower way: Han et al. (2025) target channel-permutation equivariance, Chen et al. (2025) target coordinate-based adaptation across heterogeneous devices and more than 150 layouts, and El Ouahidi et al. (2025) push setup-agnostic pretraining to more than 60,000 hours from 92 datasets and 25,000 subjects. Those are advances in recording-frame compatibility. They are still not proof that different montages, coordinate routes, and reference families already preserve one shared physiology-side representation. Ma et al. (2026) then show that strong EEG foundation models can still generalize poorly when subject-level supervision is limited unless extra adaptation structure is added, while Xiong et al. (2025), Liu et al. (2026), and Lahiri et al. (2026) show that protocol inconsistency, linear-probe versus fine-tuning regime, and pretraining-population diversity can materially change which checkpoint looks strongest.

One more front-door compression still remained: accepted conference papers, official benchmark-operation pages, and arXiv preprints could still be read as one evidence pile even though they answer different questions. The current source mix does not support that shortcut. On this site, Jiang et al. (2024) and Lee et al. (2025) are read as peer-reviewed route papers, the official EEG Challenge homepage, rules, submission page, and leaderboard are read as the current benchmark object and postmortem layer, and Han et al. (2025), Chen et al. (2025), El Ouahidi et al. (2025), Ma et al. (2026), Xiong et al. (2025), Liu et al. (2026), and Lahiri et al. (2026) are read as exploratory preprint routes or benchmark-warning analyses. That source-status split is now part of the front-door rule rather than background knowledge the reader has to infer.

The phrase harmonized EEG is still too coarse for a front-door reading. The official EEG-BIDS specification already separates electrodes, channels, coordinate system, and reference scheme, while Hu et al. (2018) showed that reference montage and electrode setup alter the measured scalp potential itself and Dong et al. (2024) validated one explicit REST-based transformation route for cross-location harmonization. Inference from these sources: common-channel intersection, interpolation to a target montage, and REST-based transformation preserve different measurement objects. So when a paper says a model `works across setups`, this FAQ still asks which recording-frame branch was used before it lets the result sound like physiology-side equivalence.

The benchmark object itself can also move. The official EEG Challenge (2025) homepage states that the proposal preprint is already out of date relative to execution-phase changes and that the current website plus Starter Kit should be treated as authoritative. The official data page shows that the benchmark family mixes six EEG tasks with psychopathology-factor prediction, while the official rules and submission page fix downsampled 100 Hz data, disclosure of extra pretraining corpora / pretrained checkpoints / fine-tuning method, a single-GPU 20 GB inference-stage constraint, and an inference-only code-submission regime. The final leaderboard then disclosed that Challenge 2 samples had not been randomized, allowing contiguous-trial same-subject structure to influence ranking and forcing separate awards. That means benchmark governance is not administrative detail; it changes what the score is allowed to mean.

At Mind-Upload, a foundation-model or leaderboard result is not promoted beyond qualified representation-learning / transfer evidence unless the paper also fixes pretraining corpus identity and overlap audit, benchmark object / supervision unit, coordinate route, reference family, omitted-channel policy, harmonization branch, adaptation regime or label budget, benchmark provenance / governance, and shortcut resistance against subject or setup fingerprints. The shortest follow-up is EEG 101, Verification: Pretraining Card, and Verification: Specificity & Shortcut Card.

If you want the site-wide reading rule for source type and status labels, continue to Wiki: How to read source types, status labels, and evidence classes.

Q. If BOLD or fNIRS changes, does that mean neural state changed?

A. Not automatically. Hemodynamic modalities carry both neural-side uncertainty and a vascular transfer state. So a group difference or longitudinal BOLD / HbO / HbR change can partly reflect baseline vascular state, cerebrovascular reactivity, or superficial/systemic contamination rather than a clean neural difference.

Murphy et al. (2011) showed that accounting for individual vascular reactivity improves group-level BOLD analyses, Williams et al. (2023) showed that task BOLD magnitude is strongly predicted by CVR across the cortex, Yücel et al. (2015) showed that short-separation regression is needed to reduce superficial confounds in fNIRS, and Epp et al. (2025) showed that significant task BOLD changes can coexist with opposite oxygen-metabolism changes in many voxels.

At Mind-Upload, a BOLD or fNIRS difference without vascular-state / CVR or short-separation audit stays a hemodynamic-limited difference rather than a clean neural difference. The shortest follow-up is Wiki: Observability and Claim Ceiling by Measurement Stack plus Verification: Observability Budget.

Q. If a paper shows EEG / MEG connectivity or information flow, did it identify communication channels or causality?

A. Not by default. On this site, sensor-space or source-space connectivity results are first read as dependence patterns under a named pipeline, not as leak-proof inter-areal communication maps or causal wiring. Vinck et al. (2011) made wPLI more conservative against some zero-lag mixing than older phase-synchrony measures, but that is not the same as eliminating leakage or proving directional influence.

The deeper problem is that the remaining failure modes are different from ordinary cleanup. Haufe et al. (2013) showed that sensor-space EEG connectivity remains strongly limited by volume conduction, Palva et al. (2018) showed that even leakage-insensitive source-space measures can generate ghost interactions, Ye et al. (2020) evaluated symbolic transfer entropy with TMS precisely because observational data alone make causality difficult to identify, and Miljevic et al. (2025) showed that sensor-space functional-connectivity estimates still change materially with rereferencing, epoch length, epoch count, and metric choice.

At Mind-Upload, a connectivity paper must therefore still disclose reference scheme and preprocessing route, leakage / volume-conduction countermeasures, source-model assumptions if used, perturbation or external validation route for directional claims, pipeline-sensitivity checks, and residual claim ceiling. If those are missing, we read the result as a pipeline-conditioned dependence map, not as a communication graph or causal circuit. The shortest follow-up is EEG 101, especially the connectivity-ceiling note, then Wiki: From observation to estimation.

Q. If a paper reports DCM or effective connectivity, did it find the brain's true causal wiring?

A. Not by default. On this site, DCM / effective-connectivity output is read as a model-conditioned causal hypothesis, not as automatic discovery of the one true circuit. Penny et al. (2004) showed that DCM conclusions are relative to the models being compared, and Rosa et al. (2012) showed that later workflows can search very large model spaces efficiently from one full model rather than make the true model unique. More recent scaling work such as Frässle et al. (2021) and Wu et al. (2024) pushes effective-connectivity estimation toward whole-brain or faster settings, but it still remains inside an explicitly chosen generative model and observation model.

That still leaves several failure modes open. Villaverde et al. (2019) showed that unknown inputs, states, and parameters often have to be assessed jointly, not one by one. Smith et al. (2011) showed that lag-based fMRI approaches perform poorly and that functionally inaccurate ROIs are extremely damaging to network estimation. Zhang et al. (2024) then showed that reasonable task-fMRI processing choices, especially GLM design and activation contrast, can materially alter group-averaged effective-connectivity patterns and parameter certainty. Barnett & Seth (2017) showed detectability black spots under subsampling, Vink et al. (2020) showed that resting-state EEG functional connectivity explains less than 10% of TMS-evoked propagation variance, Novelli et al. (2025) showed that slow BOLD sampling can still induce spurious Granger-causal inference even when realistic HRF variability alone need not do so, and Yan et al. (2026) showed that latent confounders remain an active challenge in biological network reconstruction.

That does not make effective-connectivity work useless. It means the safe claim is narrower: under a disclosed node set, omitted alternatives, observed-subsystem boundary, node-definition policy, processing / first-level design policy, prior family, and observation assumptions, one model family explained the data better than named competitors. Even reproducibility is conditional. Frässle et al. (2016), Jafarian et al. (2024), and Ma et al. (2024) show that effective-connectivity reliability can look strong under tightly matched task/rest, session-interval, scan-duration, and sample-size conditions, but that is still not the same as proving a unique causal circuit in the wild.

At Mind-Upload, a paper that says “effective connectivity” must therefore still disclose candidate model space, observed-subsystem closure / latent-confound audit, node-definition policy, processing / first-level design policy, sampling / transformation sensitivity, family comparison or model recovery, held-out perturbation / external validation, reliability window, and abstention boundary. If those are missing, we read it as a model-conditioned causal hypothesis, not as causal-wiring discovery. The shortest follow-up is Wiki: effective-connectivity route card, Verification: Observability Budget, and Roadmap R4.

Q. If a paper reports entropy production, irreversibility, or arrow-of-time in brain data, did it measure the brain's physical dissipation or WBE-ready thermodynamic cost?

A. Not by default. On this site, thermodynamic language is split before interpretation. Bérut et al. (2012) tested the Landauer lower bound for bit erasure, Attwell & Laughlin (2001) summarized tissue-side signaling costs, and papers such as Lynn et al. (2021) and Ishihara & Shimazaki (2025) estimate nonequilibrium quantities from neural data. Those objects are related, but they are not the same measurement.

In the current primary literature, the same thermodynamic vocabulary still hides different estimator families. Lynn et al. (2021) estimated entropy-production lower bounds from coarse-grained BOLD state transitions, Deco et al. (2022) used time-shifted correlation asymmetry, de la Fuente et al. (2023) used inversion decoding on ECoG, Nartallo-Kaluarachchi et al. (2025) used directed visibility graphs on MEG, and Ishihara & Shimazaki (2025) estimated model-based entropy flow under a state-space kinetic Ising model. So the phrase thermodynamic result does not by itself tell you whether the paper reported a lower bound, an asymmetry score, a graph index, or a model-based flow estimate.

Just naming the estimator family is still too weak. Lynn et al. (2021) showed that the estimate depends on how macrostates are coarse-grained, de la Fuente et al. (2023) showed that reversibility detection depends on principal-component choice, input features, and model complexity, Martínez et al. (2019) showed that waiting-time asymmetry can reveal hidden dissipation even when observable current vanishes, Blom et al. (2024) showed that coarse lumping can hide dissipative cycles and introduce memory so that entropy-production estimates become far too small when the observed trajectory is naively treated as Markov, Ishihara & Shimazaki (2025) explicitly notes that pairwise and conditional-independence assumptions limit interpretation and uses trial-shuffled controls to separate coupling-related contributions, and Baiesi et al. (2024) showed that sparse or unobserved reverse transitions can break direct entropy-production estimation even when the analysis otherwise looks clean. The next weakness is operational rather than mathematical. Poudel et al. (2024) showed that small motion can materially alter visibility-graph structure and that only low-motion subsets reached moderate-to-high test-retest reliability for selected metrics, Metzen et al. (2024) showed that BOLD variability and complexity measures have markedly different reliability profiles, and Chen et al. (2025) showed with simultaneous EEG-PET-MRI that temporal coupling across modalities can coexist with non-identical spatial organization and state trajectories. At Mind-Upload, such a paper must therefore disclose signal route and state definition, coarse-graining / timescale, observed-state closure / hidden-degree risk, estimator family and dynamical assumptions, null / surrogate control, stability / nuisance sensitivity, cross-estimator concordance if the claim is strengthened, reverse-transition support / finite-data handling, quantity type, physiology-side grounding and bridge quality if energetic language is used, cost isolation, and abstention boundary. If those are missing, we keep the result at the level of an exploratory auxiliary analysis, not a direct readout of microscopic dissipation, metabolic cost, or WBE-relevant validity. The shortest follow-up is Wiki: irreversibility route card plus Verification: thermodynamic indicators.

Q. What is the difference between decode and emulate?

A. Decode means translating observations. Emulate means having an internal state that evolves over time, responds to intervention, and generates outputs. To move closer to WBE, the benchmark has to evaluate the second kind of claim, not only the first.

Recent non-invasive word decoding and streaming speech neuroprostheses are major advances as communication routes. But at Mind-Upload, we do not read them as emulation, let alone as WBE evidence, unless the work also establishes neural contribution beyond language prior, OOD and cross-day generalization, matching after intervention, tail latency, silence, and recalibration burden, and auditing of hidden state. What has advanced first is decode, or at most local subsystem closed loop.

WBE 101, the Glossary, and Wiki: Decode vs. Emulate are the shortest follow-up route.

Q. If an LLM or digital twin talks like a person, is that Mind-Upload?

A. Not by itself. Natural conversation could reflect imitation of outward behavior, or it could reflect continuity of internal state and causal structure. Those are different questions.

Mind-Upload cares not only about whether something looks human-like, but whether the responses under changed conditions, the continuity of memory and learning, and the response to falsification conditions are all disclosed. Natural appearance matters as a clue, but it is not enough to move to an L4 identity claim.

Wiki: Counterfactual, intervention, and perturbation verification explains in stages why “it can talk naturally” is not enough.

Q. If we know the connectome and cell type, is the rest mostly filled in?

A. Not yet. Wiring and cell type still leave broad families of state latent. Gouwens et al. (2021) showed morpho-electric spread even within the same transcriptomic type, Grubb & Burrone (2010) showed activity-dependent AIS relocation that retunes excitability, Santoni et al. (2024) showed that chromatin plasticity can predetermine neuronal eligibility for memory trace formation, and Wang et al. (2015), Dai et al. (2019), Shi et al. (2018), Peterson et al. (2025), and Li et al. (2025) show that post-transcriptional RNA-state is not one row: Wang et al. (2015) is a splice-isoform route whose downstream object is chromatin / transcriptional control, Dai et al. (2019) is a splice-dependent transsynaptic receptor-balance route, Shi et al. (2018) and Li et al. (2025) are different m6A translation-versus-degradation routes, and Peterson et al. (2025) is an RNA-editing route for homeostatic AMPAR composition. Therefore, even when gene-level abundance looks similar, the operative RNA controller can still differ across memory-relevant mechanism families. Giese et al. (1998), Lee et al. (2003), Rodrigues et al. (2004), Tomita et al. (2005), and Vierra et al. (2023) show that phospho-signaling / second-messenger state is another layer: phosphosite occupancy and signaling nanodomains can still change plasticity expression even when transcript or bulk protein abundance looks similar. Govindarajan et al. (2011) showed branch-level protein-synthesis-dependent LTP integration, Park et al. (2006), Zhao et al. (2020), and Swarnkar et al. (2021) showed that compartment-specific cargo delivery is another state layer for spine growth, synaptic plasticity, and memory, Frischknecht et al. (2009) showed that ECM constrains AMPA-receptor mobility and short-term plasticity, Glykys et al. (2014) showed that local impermeant anions constrain neuronal chloride concentration, and Seidl et al. (2015) showed that node and internode geometry can tune conduction timing.

If by “connectome” one means a human diffusion-MRI tractography connectome, the object is even coarser than the word suggests, and it is not even one stable graph by default. Thomas et al. (2014) found no high-anatomical-accuracy solution across tractography methods even with exceptional ex vivo macaque diffusion data, Reveley et al. (2015) showed that superficial white matter can block long-range tracking from large parts of cortex, Donahue et al. (2016) found useful but incomplete prediction of tracer-weighted corticocortical connectivity, Schilling et al. (2020) showed that high accuracy mainly appears when strong anatomical start / end / exclusion priors are supplied, and Grisot et al. (2021) localized recurring same-brain errors at branching and turning configurations. Newer route-audit work tightened the ceiling further: Gajwani et al. (2023) showed that hub location and node strength can vary strongly across tractography pipelines and parcellations, He et al. (2024) showed that tractogram filtering can shift laterality indices for more than 10% of connections, McMaster et al. (2025) showed that voxel-size variance changes the resulting connectome, Bramati et al. (2026) showed on the same 3 T scanner with uniform processing that diffusion-sampling scheme alone still shifts voxel metrics and tractography outputs, Manzano-Patrón et al. (2025) made fibre-orientation uncertainty explicit instead of silent, and Zhu et al. (2025) improved reconstruction only by adding microscopy to MRI rather than by assuming MRI alone had already fixed the graph. At Mind-Upload, a tractography-derived human connectome is therefore read as an acquisition-, endpoint-, graph-construction-, and uncertainty-conditioned macro pathway prior or bundle-level hypothesis, not as an edge-complete human connectome. The longer rule is in Wiki: tractography route card and Verification: Observability Budget.

Likewise, Hengen et al. (2016), Torrado Pacheco et al. (2021), and Xu et al. (2024) show that sleep-dependent homeostasis and network recovery remain additional variables. Gibson et al. (2014), McKenzie et al. (2014), and Looser et al. (2024) show that myelin and oligodendrocyte support affect timing and axonal health. Separately, Hardingham & Larkman (1998), Volgushev et al. (2000), Moser et al. (1993), and Long & Fee (2008) show that local thermal-state can change synaptic reliability, spike generation, field-potential amplitude, and sequence timing even without rewiring. The human lane is already split further: healthy-human MRS thermometry (Rzechorzek et al., 2022), task-linked thermal mapping (Rogala et al., 2024), and frontal-lobe thermometry (Tan et al., 2025) remain passive or task-linked macro thermal proxies, while Tan et al. (2024) and Inoue et al. (2025) add bounded human perturbation-conditioned thermal routes through severe heat exposure and intraoperative focal cooling. None of those human rows, however, is cell-specific thermal ground truth. Separately, Rangaraju et al. (2014), Rangaraju et al. (2019), Divakaruni et al. (2018), Bapat et al. (2024), and Hu et al. (2025) show that local ATP supply, mitochondrial positioning, fission/fusion, and synaptic ATP-synthase organization remain additional hidden state for repeated-burst reliability and dendritic plasticity. Current human energetic routes now have to be split more narrowly: Ren et al. (2015) is a 31P metabolite / pH balance route, Ren et al. (2017) is a 31P MT exchange-flux route, Guo et al. (2024) is a whole-brain NAD-content route, Kaiser et al. (2026) is a functionally localized NAD-dynamics route, Karkouri et al. (2026) is a deuterium metabolite-mapping / absolute-quantification route, and Li et al. (2025) is a deuterium kinetic-rate route. Those human advances are real, but they are still macro energetic proxies rather than branch-local mitochondrial readouts, and they are not interchangeable with one another. The same human ceiling applies to post-transcriptional RNA-state: current in vivo routes and ordinary short-read transcript summaries do not directly reveal isoform choice, m6A reader engagement, or RNA-editing ratios across the whole living human brain, while specialized long-read atlas work such as Joglekar et al. (2024) remains a targeted ex vivo / atlas-building route rather than a whole-brain in vivo readout. The same human ceiling also applies to phospho-signaling / second-messenger state: current in vivo routes do not directly reveal phosphosite occupancy, kinase/phosphatase balance, or compartment-specific signaling nanodomains across the living whole human brain, while ex vivo phosphoproteome atlas work such as Biswas et al. (2023) remains an atlas route rather than a living whole-brain readout. The same human ceiling also applies to cargo routing: current in vivo routes do not directly reveal branch-specific motor engagement, cargo pausing, or bouton-level retention. Likewise, human sodium MRI routes such as Qian et al. (2012) and Qian et al. (2025) are important macro ionic proxies, but they still do not directly reveal cell-specific chloride concentration, KCC2 / NKCC1 balance, or local E_GABA. The same split now applies to glial metabolism / substrate routing: Suzuki et al. (2011) is a lactate-shuttle route, Silva et al. (2022) is a starvation ketone-body route, Pavlowsky et al. (2025) is an intensive-learning glia-to-neuron fatty-acid route, and Greda et al. (2025) is an apoE / sortilin-dependent lipid-delivery and fuel-choice route. These papers do not share one fuel object, one supplier cell, one neuronal sink, one regime trigger, or one human observability ceiling, so this site does not treat `glial support` as one solved row. Astrocyte-state then stays separate again from glial fuel routing: Cahill et al. (2024), Williamson et al. (2025), Dewa et al. (2025), and Bukalo et al. (2026) show local transmitter encoding, recall, multiday stabilization, and fear-state representation routes, while human astrocyte PET routes such as Villemagne et al. (2022), Tyacke et al. (2018), and Livingston et al. (2022) remain target-defined proxy families rather than local controller readouts. On this site, those rows raise the need to keep both glial substrate-routing and astrocyte-state explicit, but they still do not permit a jump to direct human whole-brain memory readout.

The same bucket is too coarse for sleep architecture / replay-coupling. Ngo et al. (2013) showed that auditory stimulation benefits memory only under a phase-locked slow-oscillation policy, Baxter et al. (2023) showed that closed-loop stimulation can alter SO / spindle dynamics while still failing to improve memory if sleep continuity is disturbed, Whitmore et al. (2022) showed that TMR benefit depends on ample and undisturbed N3 sleep, Schreiner et al. (2021) constrained endogenous scalp-EEG decoding around aggregated slow-oscillation / spindle events, Schreiner et al. (2023) showed that respiration-linked SO-spindle coupling is related to reactivation strength, Geva-Sagiv et al. (2023) showed that an intracranial closed-loop synchrony intervention improves human overnight memory only when precisely time-locked, Schreiner et al. (2024) linked spindle-locked ripples to human memory reactivation, Whitmore et al. (2024) showed that sleep-disruption effects depend on memory age, Jourde et al. (2025) showed that the effectiveness of auditory stimulation depends on thalamocortical spindle phase, Duan et al. (2025) showed that one human TMR session can contain both strengthening and decaying items, Shin et al. (2025) showed that behavioral benefit can concentrate in challenging memories rather than across all items, and Deng et al. (2025) showed that even within NREM the consolidation window is time-structured. At Mind-Upload, that means sleep happened, oscillations increased, a cue was delivered, overnight memory changed, and replay-coupling matched are not treated as interchangeable statements. The route has to log sleep-integrity burden, NREM physiology gating, and memory age / selection regime in addition to timing. The longer rule is in Wiki: sleep replay route card and Verification: maintenance-state error budget.

The same correction also applies to neurovascular-unit / BBB / pericyte state. Bell et al. (2010) showed that adult pericyte loss produces hypoperfusion, BBB breakdown, and later memory impairment, Kisler et al. (2020) showed rapid neurovascular uncoupling after acute pericyte ablation, Pandey et al. (2023) showed that pericyte-derived IGF2 is required for long-term memory, Swissa et al. (2024) linked cortical plasticity to BBB modulation, and Mai-Morente et al. (2025) showed that a pericyte capillary-diameter controller affects memory. Current human BBB routes such as Padrela et al. (2025) and Morgan et al. (2024) constrain water exchange, while Chung et al. (2025) constrains tracer-specific BBB transport. A distinct human blood-CSF-barrier lane also exists, and it is not one internal quantity either: Zhao et al. (2020) and Sun et al. (2024) constrain choroid-plexus perfusion, Petitclerc et al. (2021) constrains blood-to-CSF water transport, Anderson et al. (2022) constrains choroid-plexus water cycling, Wu et al. (2026) constrains apparent BCSFB exchange, and Petitclerc et al. (2026) constrains joint BBB-versus-BCSFB ASL exchange in one acquisition. Those are still bounded human support-state routes rather than direct readouts of the responsible pericyte, endothelial, or choroid-plexus epithelial controller. At Mind-Upload, this layer is therefore kept separate both from measurement-side vascular-state / CVR audit and from clearance / immune support.

The same correction applies to shared extracellular / electrical state. Galarreta & Hestrin (1999) showed that fast-spiking interneurons can form electrical-synapse networks, Anastassiou et al. (2011) showed that endogenous extracellular fields can causally bias cortical spike timing, Burman et al. (2023) showed that active cortical networks can shift fast inhibition toward a predominantly shunting regime in vivo, Yang et al. (2024) showed that activity-dependent electrical synapses can rewire local networks for persistent oscillations, and Selfe et al. (2024) showed that direct inhibitory driving-force measurement requires a specialized local optical route. Human evidence is route-split even before any strong reading: Feld et al. (2026) is useful as a perturbation-conditioned clue that electrical coupling can matter for spindle-to-slow-oscillation coordination during sleep, but it is still not a direct whole-brain readout of local electrical state. At Mind-Upload, a chemical connectome plus nominal inhibitory edges is therefore not read as electrical-state complete. If a paper mixes these routes, this site now asks for an electrical-state route card that names claim family, direct observable, spatial regime, perturbation / calibration route, human evidence class, and abstention.

The same bucket was also too coarse for extracellular-space geometry / diffusion-barrier / osmotic-regime routes. Graydon et al. (2014) showed that synapse-adjacent morphology changes extracellular dilution and signaling, Kilb et al. (2006) and Lauderdale et al. (2015) showed that osmotic ECS contraction / edema can rapidly increase excitability, Xie et al. (2013) showed sleep-linked interstitial-space expansion in mice, Voldsbekk et al. (2020) gave a bounded human diffusion-MRI clue consistent with wakefulness-related extra-axonal / extracellular-volume reduction, and Örzsik et al. (2023) added a sleep-conditioned higher-order diffusion / glymphatic clue under a within-subject sleep-deprivation-plus-zolpidem regime. Therefore, at Mind-Upload a chemical connectome plus nominal inhibitory edges is not read as extracellular-state complete either.

At Mind-Upload, this means we treat connectome-complete as progress on the structural scaffold, not as emulation-complete. Current excitability, timing-state, thermal-state, transcription/chromatin, post-transcriptional RNA-state, local proteostasis, cargo-transport / cytoskeletal trafficking state, ECM / PNN gate, ionic milieu, bioenergetic / mitochondrial state, neurovascular-unit / BBB / pericyte state, sleep/controller state, glial substrate-routing, astrocyte-state, and clearance / immune support still need to be disclosed or left explicitly latent. So same-day activity matching, cross-day stability, and maintenance-consistent dynamics remain separate claims. The shortest follow-up is Wiki: Why wiring diagrams alone are not enough plus Wiki: Homeostatic plasticity and maintenance state.

The same correction applies to clearance / immune support. Louveau et al. (2015) and Ahn et al. (2019) established meningeal-lymphatic drainage routes, Kim et al. (2025) showed that the meningeal-lymphatics-microglia axis can regulate synaptic physiology, and the human lane is no longer transport-only. Biechele et al. (2023) showed why TSPO is not a universal human activation-state meter, Wijesinghe et al. (2025) validated TSPO PET as a microglial biomarker in PSP, Horti et al. (2022) plus Ogata et al. (2025) established first-in-human CSF1R PET routes, and Yan et al. (2025) quantified COX-2 in healthy human brain. Fultz et al. (2019) then showed a macroscopic sleep-linked CSF-oscillation route, Kim, Huang, & Liu (2025) showed a parenchyma-CSF water-exchange route, Lim et al. (2025) showed an awake-state respiration-conditioned CSF net-flow route, Yoo et al. (2025) showed an exercise-conditioned contrast-influx / meningeal-lymphatic-flow route, Eide et al. (2023) showed an intrathecal tracer / CSF-to-blood clearance-capacity route, Hirschler et al. (2025) showed a CSF-mobility route, and Dagum et al. (2026) showed a model-based sleep-linked biomarker-efflux route. On this site, that means clearance is not treated as passive cleanup, but current human evidence is read as two bounded lanes, macro support-state transport proxy and target-defined neuroimmune PET, not as direct identification of a local immune controller or synapse-specific maintenance mechanism.

The separate question of what human measurement can actually observe today is handled in Q2d. That separation is deliberate: connectome insufficiency and human observability ladder are related, but they are not the same reading task.

Q. If same-brain functional connectomics or a digital twin works, did we solve a local twin?

A. Not yet. Bosch et al. (2022) showed that linking live physiology to later ultrastructure already requires a multistage landmark-based bridge, and MICrONS Consortium et al. (2025) strengthened that route to a same-brain local dataset with about 75,000 neurons linked to a later EM reconstruction of more than 200,000 cells and 0.5 billion synapses. That is a major advance in same-brain local structure-function linkage, but it is still a sequential local pipeline, not a simultaneous whole-state sample. Ding et al. (2025) then added a validated stimulus-conditioned response model, while also warning that the model's internal representations still need cautious interpretation. Gamlin et al. (2025) sharpened the cell-type bridge, but still through morphology-based predicted transcriptomic labels rather than direct transcriptomic assay inside the EM volume.

The remaining ceiling is not cosmetic. Holler et al. (2021) showed that inferring function from wiring diagrams is limited by unresolved synaptic-strength structure, Molnár et al. (2016) showed that human synapses can contain multiple docked vesicles and multivesicular release, Sakamoto et al. (2018) showed that Munc13-1 supramolecular assemblies set independent release sites, Dürst et al. (2022) showed that vesicular release probability strongly sets synaptic strength, Emperador-Melero et al. (2024) showed that Ca_V2 clustering and vesicle priming are executed by distinct active-zone machineries, and Mittermaier et al. (2024) showed that membrane-potential state gates synaptic consolidation in human neocortical tissue. Beiran & Litwin-Kumar (2025) then showed that connectome-constrained recurrent networks can still remain dynamically degenerate until additional recordings narrow the compatible family. So at Mind-Upload the safe ceiling is a sequential local structure-function scaffold plus, at most, a task-bounded conditional predictor. It is not direct transcriptomic truth, not a readout of release-site number, docked-vesicle architecture, active-zone nanostructure / priming-site assembly, or current release competence, and not one solved local twin. The shortest follow-up is Wiki: Why a Connectome Is Not Enough, Wiki: Observability and Claim Ceiling by Measurement Stack, Wiki: State-Continuity Bridge, and Wiki: Decode vs. Emulate.

Q. If human measurement keeps improving, are we close to state-complete readout?

A. Not yet. The first split is between destructive local ex vivo structure and living-human in vivo proxy routes. The second split is inside the human routes themselves: recent human advances do not all report the same quantity. Lu et al. (2023) showed why preservation route changes extracellular-space retention and native geometry, and Shapson-Coe et al. (2024) reconstructed a cubic millimeter of human temporal cortex at nanoscale resolution, but as a rapidly preserved local surgical fragment, not a living whole-brain state readout. By contrast, Johansen et al. (2024) built an SV2A PET atlas in healthy humans (17F/16M) calibrated with postmortem autoradiography, which is a regional synaptic-density proxy. Smart et al. (2021) showed that brief visual activation does not measurably change [¹¹C]UCB-J binding, and Holmes et al. (2022) found no measurable overall SV2A change 24 h after ketamine despite symptom improvement, so even a strong SV2A PET route remains a regional density proxy rather than a momentary release-machinery readout. Lucchetti et al. (2025) used five-metabolite ¹H-MRSI in 51 healthy participants with an independent replication sample of 13 to derive a parcel-level biochemical similarity scaffold, while Guo et al. (2025) used ultrahigh-field extended spatiospectral encoding plus subspace modeling to produce high-resolution ¹H-MRSI metabolite-distribution maps under explicit ghosting / aliasing / low-SNR handling. The spectroscopy rows already split further: Ren et al. (2015) reported resting ATP synthesis, phosphorus-metabolite concentrations, pH, and T₁ in 12 resting subjects; Ren et al. (2017) estimated PCr→ATP and Pi→ATP exchange fluxes from a 5-pool magnetization-transfer model in six resting subjects; Guo et al. (2024) mapped whole-brain intracellular NAD content at 2.3 cc nominal resolution with a 1.0 cc feasibility dataset; Kaiser et al. (2026) used prior fMRI to localize a visual-cortex voxel and reported task-evoked NAD⁺ dynamics in 25 healthy volunteers; Karkouri et al. (2026) used a dedicated absolute-quantification pipeline at 7 T to produce HDO / Glc / Glx / Lac maps and associated rate estimates in a mixed cohort of 12 healthy volunteers and 5 treatment-naive glioblastoma patients, with only two healthy volunteers studied after [6,6'-²H₂]glucose; Li et al. (2025) estimated glucose-transport and metabolic-rate maps using 7 T dynamic DMRSI, blood-input / kinetic modeling, and 0.7 cc nominal voxels in five healthy participants; Ahmadian et al. (2025) showed that [6,6'-²H₂]glucose dose materially changes human brain-side deuterated-glucose and Glx visibility; and Bøgh et al. (2024) showed that repeatability at 3 T depends on a named acquisition and time-point regime. van Blooijs et al. (2023) estimated tract-scale transmission speed in a developmental human route, which is a timing-support advance but not a myelin-specific MRI quantity. Baadsvik et al. (2024) demonstrated myelin-bilayer mapping in two healthy volunteers. Morgan et al. (2024) showed that even BBB water-exchange estimates differ materially across ASL route choices, while Chung et al. (2025) quantified tracer-specific BBB permeability-surface-area product with dynamic PET and kinetic modeling, but also stated that human ground truth and test-retest validation remain future work. A distinct human blood-CSF-barrier lane also exists: Zhao et al. (2020) measured choroid-plexus perfusion, Sun et al. (2024) extended that perfusion lane with a 641-person HCP-Aging analysis, Petitclerc et al. (2021) measured blood-to-CSF water transport, Anderson et al. (2022) estimated choroid-plexus water cycling, Wu et al. (2026) reported apparent BCSFB exchange with scan-rescan repeatability, and Petitclerc et al. (2026) jointly estimated BBB-versus-BCSFB ASL exchange in one acquisition. Villemagne et al. (2022) and Tyacke et al. (2018) then showed that human astrocyte-related PET already splits into MAO-B and I₂BS target classes rather than one generic astrocyte scalar. Finally, Fultz et al. (2019) showed macroscopic NREM-linked CSF oscillations, Kim, Huang, & Liu (2025) showed parenchyma-CSF water exchange in healthy humans, Eide et al. (2023) linked intrathecal gadobutrol retention and PK-based CSF-to-blood clearance variables to plasma biomarkers in neurological patients, Hirschler et al. (2025) showed whole-brain rest CSF-mobility maps in 20 analyzed healthy younger individuals together with a separate CAA comparison cohort, and Dagum et al. (2026) advanced sleep-linked biomarker efflux through an investigational device plus multicompartment model in a multi-site randomized crossover study whose primary analysis included 39 participants.

At Mind-Upload, the safe reading is now three-axis: name which proxy class and quantity type the route constrains, separately name how specialized, small-cohort, regime-locked, or model-heavy the route still is, and then state what bounded calibrator role the route safely plays. If you skip that third step, you can silently turn "real human route" into "broad hidden-state calibration". If you skip the second, you can silently turn "human evidence got richer" into "human state-complete measurement is close". If you skip the first, you can silently turn "human MRI, PET, or paired-fluid proteomics exists" into "one common latent state is already being measured". The current primary literature does not support any of those jumps.

What follows directly is that route maturity is not just sample size. It also includes whether the paper is an atlas, a single-voxel task study, a mixed pathology/healthy method paper, an exploratory clinical application, or a sleep-manipulation model study. On this site, those operating-point differences are logged before human rows are composed.

A further tightening follows on this site: even if those three axes are logged, the bundle is still not promoted unless the composition audit says whether it passes a robustness gate, a common-driver / quantity-bridge gate, and an increment-over-the-strongest-single-row gate.

The same stop line matters even before multimodal fusion. If a bundle crosses developmental stage, aging regime, or pathology enrichment, this site asks for an explicit transfer argument rather than silently importing one cohort's baseline into another.

One more correction is needed before overreading the same literature: the strongest maintenance-state causal papers and the strongest current human-observability papers are often on different ladders. Hadzibegovic et al. (2025), Terceros et al. (2026), and Bukalo et al. (2026) strengthen local causal relevance, while Zrenner et al. (2018), Hirschler et al. (2025), and Dagum et al. (2026) strengthen human perturbation or proxy routes. That combination still does not mean that the responsible human controller was measured. The short crosswalk is in WBE 101: causal relevance vs human observability.

A second safety rule follows from the same literature: local human ultrastructure is not simply the first rung of a living-human proxy ladder. On this site it is paired with the Destructive-Structure Route Card because preservation route, live-to-fix window, registration scope, and proofreading burden still change what the paper means.

That caution applies especially to human clearance and neuroimmune papers. On this site, macroscopic CSF oscillation, parenchyma-CSF water exchange, respiration-conditioned net flow, exercise-conditioned contrast influx / meningeal-lymphatic flow, intrathecal tracer / CSF-to-blood clearance, CSF-mobility MRI, and sleep-linked biomarker-efflux studies lower a support-state uncertainty, while TSPO, CSF1R, and COX-2 PET lower a target-defined neuroimmune uncertainty. But none of those rows tells you which meningeal-lymphatic segment, which microglial controller, or which local synapse is responsible for the difference. They therefore stay below a local maintenance-controller claim ceiling.

The shortest follow-up is WBE 101: human observability ladder, Wiki: Human Proxy Composition and Route Maturity, and Wiki: Observability and Claim Ceiling by Measurement Stack. Those pages keep proxy class, route maturity, composition failure modes, and the still-latent hidden-state families in one place. If the argument then promotes same-subject or same-brain wording into one state sample across stages or days, continue directly to Q2e.

Q. If a paper says same-subject or same-brain, does that mean one state was captured?

A. Not automatically. On this site, same-subject or same-brain solves specimen identity, not state continuity. A sequential bridge still has to justify what object / witness is supposed to survive, how failure would have been detected, elapsed time, physiological-regime continuity, coordinate transfer, and bridge validation before one later stage is read as the same state rather than the same specimen at another stage.

The primary literature is already strong enough to require that separation. Lu et al. (2023) showed that preservation route changes extracellular-space retention and native geometry, and Idziak et al. (2023) showed subtle spine-morphology changes plus substantial membrane damage after chemical fixation. Bosch et al. (2022) and MICrONS Consortium et al. (2025) then showed why even very strong correlative same-brain workflows remain multistage local bridges with landmarks, targeted subvolumes, and later reconstruction rather than simultaneous whole-state capture. In other words, the bridge may carry a local correspondence object without carrying one global latent state.

The same warning applies even without fixation. Gallego et al. (2020) showed that a latent manifold can remain stable despite turnover in recorded neurons, Noda et al. (2025) showed recovery of a population-level representational map after selective neuron loss, and Van De Ville et al. (2021) plus Di et al. (2021) showed that fingerprint-like identifiability depends on timescale and feature family. Those are different carried objects, not interchangeable proof that the same full state survived.

Stable performance is also not the same thing as raw continuity. Karpowicz et al. (2025) stabilized decoding by aligning recordings to a consistent latent-dynamics object, Wilson et al. (2025) maintained long-term cursor control through repeated unsupervised recalibration, and Wairagkar et al. (2025) reported a clinically important speech neuroprosthesis while still relying on a bounded fixed-decoder horizon. Benisty et al. (2024) and Egger et al. (2024) further showed that spontaneous behavior and 10-hour EEG dynamics can move the measured state within hours. Therefore, a cross-day or cross-regime bridge still cannot be read as one stable state sample by wording or score alone.

At Mind-Upload, if those fields are missing, we read the result as specimen linkage only, witness-specific continuity only, or at most a sequential local scaffold, not as same-state evidence. The shortest follow-up is Verification: State-Continuity Bridge Card and Wiki: State-Continuity Bridge. If the bridge crosses hours to days in a live stack, add Verification: Temporal Validity Card as well.

Q. Then what should we build to count as progress?

A. For now, L0-L2 is the realistic target: reproducible analyses, comparable benchmarks, and models that can be tested through intervention prediction. Mind-Upload turns those into operating templates, logs, and rules that the site can actually use.

If you want to see how those three parts connect in a single EEG example, the shortest route is Wiki: Verification example walkthrough.

Q. Why is standardization so important?

A. Without standards, people appear to be talking about the same thing while actually comparing different inputs, different procedures, and different metrics. Once that happens, progress becomes unreadable.

Examples such as the PDB and the BIDS + OpenNeuro ecosystem differ in field, but they share the same crucial property: they turned progress into something different groups could measure in comparable ways. The casework collection summarizes the design pattern, and if you want the role differences among BIDS, OpenNeuro, validators, and benchmarks first, the shortest path is Wiki: Standards, repositories, validators, and benchmarks.

Q. What is the “benchmark trap”?

A. It is the phenomenon where winning on a metric drifts away from the real goal, a version of Goodhart's law. For example, the score may go up because of data leakage or overfitting, or the system may be too costly to deploy in practice. Mind-Upload treats failure cases, leakage checks, and model cards as part of the benchmark design itself.

Wiki: Dataset splits and data leakage collects typical ways the numbers break when train/test separation is poorly designed.

Q. If offline accuracy is high, is that enough for closed loop?

A. No. In a closed loop, the output changes the next input and often the environment as well, so end-to-end latency, jitter, drift, and safe-stop behavior all matter. A method can work well on recorded data and still fail to operate stably in real time.

Recent speech neuroprosthesis work made major progress in real-time text, audio, and voice output. Littlejohn et al. (2025) showed streaming brain-to-voice output every 80 ms, and Wairagkar et al. (2025) showed neural-to-voice synthesis under 10 ms. But those are within-session achievements in invasive communication subsystems, not general evidence for WBE. As Wilson et al. (2025) shows, whether such systems can be maintained long-term while reducing daily supervised recalibration is a separate question.

At Mind-Upload, offline accuracy and L3 closed-loop stability are read separately. In particular, if the work does not report P50/P95/P99 latency, silence or abstention behavior, recalibration burden, and cross-day degradation apart from accuracy, we do not read it as a deployable closed loop. For a beginner-friendly version of that distinction, see Wiki: Closed loop, latency, jitter, and safe stops.

The same front-door caution now applies to adaptive DBS and other burst- or state-triggered neuromodulation. Recent primary literature does not support reading them as one generic `adaptive` success. Beta-guided bradykinesia control, gamma-linked dyskinesia or prokinetic routes, and decoder-based movement-responsive control are different controller families, and movement state, medication state, sensing compatibility, comparator policy, and programming burden can all change what the result means. So on this site, a positive symptom result is not read as generic proof that state-dependent control has been solved unless those controller details are visible.

Q. If latency is low, does that mean the body/environment problem is solved?

A. No. Low latency only tells you that one loop is fast enough; it does not tell you that the relevant subject boundary has been reproduced. Musall et al. (2019) and Stringer et al. (2019) showed that ongoing behavior shapes a large fraction of cortical activity, Saleem et al. (2013) and Ravassard et al. (2013) showed that locomotion, optic flow, vestibular, and other sensory cues reshape cortical and hippocampal codes, Zelano et al. (2016) showed that nasal respiration entrains human limbic activity and modulates memory, and Flesher et al. (2021) showed that adding tactile feedback improves a local bidirectional BCI loop.

That fast-loop disclosure is still not enough by itself. de Quervain et al. (1998) and Oei et al. (2007) showed that glucocorticoid state can impair retrieval and reduce hippocampal / prefrontal retrieval activity, while McCauley et al. (2020), Barone et al. (2023), and Birnie et al. (2023) showed that circadian timing and corticosteroid rhythm alter hippocampal plasticity. At Mind-Upload, a closed-loop paper that does not disclose which sensory, motor, interoceptive, and slow internal-milieu routes were preserved, substituted, matched, perturbed, or omitted, plus what happened when those loops were removed or altered, stays a local controller or local subsystem-loop result. The shortest follow-up is Verification: Body / Environment Boundary Card together with Wiki: before milliseconds, fix which loop boundary was actually preserved, Roadmap M4, and Roadmap I6.

Q. What is the site's stance on the hard problem of consciousness?

A. Mind-Upload does not assume any specific philosophical stance on the hard problem of consciousness (Chalmers, 1995). It uses a functionalist approach as an implementation basis, but it does not claim that functional equivalence is sufficient for phenomenal consciousness. Instead, it evaluates verifiable operational indicators such as PCI-ST, counterfactual tests, and agreement under intervention, and leaves philosophical consequences to the interpretation stage once there is enough empirical data.

Q. How do you handle the copy problem?

A. In a scan-and-copy route, the original and the copy branch immediately after the copy is made, so there is no principled way to decide which one is “the person” in the strong sense associated with identity debates. Mind-Upload therefore centers slow continuous mind uploading as a design strategy, treating continuity of process without abrupt rupture as an engineering requirement.

Even then, the strategy still needs a pre-fixed criterion for when the transfer is complete. Mind-Upload treats that as part of the L4 claim ladder rather than as something already solved.

Q. What happened with experimental tests of IIT and GNWT?

A. The 2025 Cogitate Consortium adversarial collaboration tested predictions from IIT and GNWT at large scale. The result was that neither theory received full support: IIT's posterior cortical sustained activity received only partial support, while GNWT's prefrontal ignition remained difficult to separate from report-related activity. In response, Mind-Upload avoids overcommitting to any one theory and instead emphasizes theory-light empirical indicators such as PCI.

If you want the theory map itself first, Wiki: Consciousness theory map is the shortest route.

Q. How does the site treat ethical issues?

A. WBE brings its own ethical issues, including (1) the legal status and rights of emulated beings, (2) whether consent can be withdrawn, including the right to stop, (3) the ethics of multiple copies, and (4) access inequality and social justice. Mind-Upload places the design of ethical review and governance as a prerequisite for L5 social deployment while focusing current effort on the technical foundation in L0-L2.

Q. How is this different from other WBE-related projects?

A. Relative to several major existing efforts, the rough distinction is:

• Blue Brain / Human Brain Project: more focused on large-scale simulation. Mind-Upload differs by prioritizing the verification framework first.
• Whole Brain Architecture Initiative (WBAI): more focused on constructive architecture and roadmap work. Mind-Upload is complementary in that it fixes benchmarks and falsification conditions early.
• OpenWorm: focused on full-connectome implementation for C. elegans. Mind-Upload instead starts from non-invasive human-brain measurement, especially EEG.

The distinguishing strategy of Mind-Upload is to build the Verification Commons first.

References (FAQ)

1. Tang, J., et al. (2023). Semantic reconstruction from non-invasive brain recordings. doi:10.1038/s41593-023-01304-9
2. Défossez, A., Caucheteux, C., Rapin, J., et al. (2023). Decoding speech perception from non-invasive brain recordings. doi:10.1038/s42256-023-00714-5
3. d'Ascoli, S., Bel, C., Rapin, J., et al. (2025). Towards decoding individual words from non-invasive brain recordings. doi:10.1038/s41467-025-65499-0
4. Ye, Z., Ai, Q., Liu, Y., de Rijke, M., Zhang, M., Lioma, C., & Ruotsalo, T. (2025). Generative language reconstruction from brain recordings. doi:10.1038/s42003-025-07731-7
5. Rybár, M., Poli, R., & Daly, I. (2024). Using data from cue presentations results in grossly overestimating semantic BCI performance. doi:10.1038/s41598-024-79309-y
6. Horikawa, T. (2025). Mind captioning: Evolving descriptive text of mental content from human brain activity. doi:10.1126/sciadv.adw1464
7. Unnwongse, K., Achakulvisut, T., Wu, J. Y., et al. (2023). Validating EEG source imaging using intracranial electrical stimulation in focal epilepsy. doi:10.1093/braincomms/fcad023
8. Hao, S., Zhao, H., Feng, Z., et al. (2025). HD-EEG source imaging with simultaneous SEEG recording in drug-resistant epilepsy. doi:10.1111/epi.18552
9. Mahjoory, K., Nikulin, V. V., Botrel, L., et al. (2017). Consistency of EEG source localization and connectivity estimates. doi:10.1016/j.neuroimage.2017.02.076
10. Mikulan, E., Russo, S., Parmigiani, S., et al. (2020). Simultaneous human intracerebral stimulation and HD-EEG, ground-truth for source localization methods. doi:10.1038/s41597-020-0467-x
11. Ahlfors, S. P., Han, J., Belliveau, J. W., & Hämäläinen, M. S. (2010). Sensitivity of MEG and EEG to source orientation. doi:10.1007/s10548-010-0154-x
12. Ahlfors, S. P., Han, J., Lin, F.-H., Witzel, T., Belliveau, J. W., Hämäläinen, M. S., & Halgren, E. (2010). Cancellation of EEG and MEG signals generated by extended and distributed sources. doi:10.1002/hbm.20851
13. Goldenholz, D. M., Ahlfors, S. P., Hämäläinen, M. S., Sharon, D., Ishitobi, M., Vaina, L. M., & Stufflebeam, S. M. (2009). Mapping the signal-to-noise-ratios of cortical sources in magnetoencephalography and electroencephalography. doi:10.1002/hbm.20571
14. Piastra, M. C., Nüßing, A., Vorwerk, J., Clerc, M., Engwer, C., & Wolters, C. H. (2021). A comprehensive study on electroencephalography and magnetoencephalography sensitivity to cortical and subcortical sources. doi:10.1002/hbm.25272
15. Willett, F. R., Kunz, E. M., Fan, C., et al. (2023). A high-performance speech neuroprosthesis. doi:10.1038/s41586-023-06377-x
16. Littlejohn, K. T., Dabagia, M., Ladwig, A., et al. (2025). A streaming brain-to-voice neuroprosthesis to restore naturalistic communication. doi:10.1038/s41593-025-01905-6
17. Wairagkar, M., Card, N. S., Singer-Clark, T., et al. (2025). An instantaneous voice-synthesis neuroprosthesis. doi:10.1038/s41586-025-09127-3
18. Singh, A., Wu, E., Ramsey, N. F., et al. (2025). Transfer learning via distributed brain recordings enables reliable speech decoding. doi:10.1038/s41467-025-63825-0
19. Karpowicz, B. M., Ali, Y. H., Wimalasena, L. N., et al. (2025). Stabilizing brain-computer interfaces through alignment of latent dynamics. doi:10.1038/s41467-025-59652-y
20. Wilson, G. H., Stein, E. A., Kamdar, F., et al. (2025). Long-term unsupervised recalibration of cursor-based intracortical brain-computer interfaces using a hidden Markov model. doi:10.1038/s41551-025-01536-z
21. Penny, W. D., Stephan, K. E., Mechelli, A., & Friston, K. J. (2004). Comparing dynamic causal models. doi:10.1016/j.neuroimage.2004.03.026
22. Rosa, M. J., Friston, K., & Penny, W. (2012). Post-hoc selection of dynamic causal models. doi:10.1016/j.jneumeth.2012.04.013
23. Frässle, S., Paulus, F. M., Krach, S., & Jansen, A. (2016). Test-retest reliability of effective connectivity in the face perception network. doi:10.1002/hbm.23061
24. Frässle, S., Manjaly, Z. M., Do, C. T., Kasper, L., Pruessmann, K. P., & Stephan, K. E. (2021). Whole-brain estimates of directed connectivity for human connectomics. doi:10.1016/j.neuroimage.2020.117491
25. Wu, H., Hu, X., & Zeng, Y. (2024). A fast dynamic causal modeling regression method for fMRI. doi:10.1016/j.neuroimage.2024.120954
26. Almgren, H., Van de Steen, F., Razi, A., Friston, K., & Marinazzo, D. (2020). The effect of global signal regression on DCM estimates of noise and effective connectivity from resting state fMRI. doi:10.1016/j.neuroimage.2019.116435
27. Zhang, S., Jung, K., Langner, R., Florin, E., Eickhoff, S. B., & Popovych, O. V. (2024). Impact of data processing varieties on DCM estimates of effective connectivity from task-fMRI. doi:10.1002/hbm.26751
28. Ma, L., Braun, S. E., Steinberg, J. L., Bjork, J. M., Martin, C. E., Keen II, L. D., & Moeller, F. G. (2024). Effect of scanning duration and sample size on reliability in resting state fMRI dynamic causal modeling analysis. doi:10.1016/j.neuroimage.2024.120604
29. Jafarian, A., Assem, M. K., Kocagoncu, E., et al. (2024). Reliability of dynamic causal modelling of resting-state magnetoencephalography. doi:10.1002/hbm.26782
30. Gouwens, N. W., et al. (2021). Integrated morphoelectric and transcriptomic classification of cortical GABAergic cells. doi:10.1038/s41586-020-2907-3
31. Hengen, K. B., Torrado Pacheco, A., McGregor, J. N., Van Hooser, S. D., & Turrigiano, G. G. (2016). Neuronal firing rate homeostasis is inhibited by sleep and promoted by wake. doi:10.1016/j.cell.2016.01.046
32. Torrado Pacheco, A., et al. (2021). Sleep Promotes Downward Firing Rate Homeostasis. doi:10.1016/j.neuron.2021.04.004
33. Xu, W., et al. (2024). Sleep restores an optimal computational regime in cortical networks. doi:10.1038/s41467-024-47838-5
34. Ngo, H.-V. V., Martinetz, T., Born, J., & Mölle, M. (2013). Auditory closed-loop stimulation of the sleep slow oscillation enhances memory. doi:10.1016/j.neuron.2013.03.006
35. Whitmore, N. W., Bassard, A. M., & Paller, K. A. (2022). Targeted memory reactivation of face-name learning depends on ample and undisturbed slow-wave sleep. npj Science of Learning, 7, 1. doi:10.1038/s41539-021-00119-2
36. Baxter, B. S., Mylonas, D., Kwok, K. S., Talbot, C. E., Patel, R., Zhu, L., Vangel, M., Stickgold, R., & Manoach, D. S. (2023). The effects of closed-loop auditory stimulation on sleep oscillatory dynamics in relation to motor procedural memory consolidation. Sleep, 46(10), zsad206. doi:10.1093/sleep/zsad206
37. Schreiner, T., Petzka, M., Staudigl, T., & Staresina, B. P. (2021). Endogenous memory reactivation during sleep in humans is clocked by slow oscillation-spindle complexes. doi:10.1038/s41467-021-23520-2
38. Schreiner, T., Petzka, M., Staudigl, T., et al. (2023). Respiration modulates sleep oscillations and memory reactivation in humans. Nature Communications, 14, 8351. doi:10.1038/s41467-023-43450-5
39. Geva-Sagiv, M., Mankin, E. A., Eliashiv, D., et al. (2023). Augmenting hippocampal-prefrontal neuronal synchrony during sleep enhances memory consolidation in humans. doi:10.1038/s41593-023-01324-5
40. Schreiner, T., Griffiths, B. J., Kutlu, M., et al. (2024). Spindle-locked ripples mediate memory reactivation during human NREM sleep. doi:10.1038/s41467-024-49572-8
41. Whitmore, N. W., Yamazaki, E. M., & Paller, K. A. (2024). Targeted memory reactivation with sleep disruption does not weaken week-old memories. npj Science of Learning, 9, 64. doi:10.1038/s41539-024-00276-0
42. Jourde, N., Fattinger, S., Teague, M., et al. (2025). The effectiveness of auditory stimulation in sleep varies with thalamocortical spindle phase. doi:10.1016/j.neuroimage.2025.121530
43. Duan, W., Xu, Z., Chen, D., et al. (2025). Electrophysiological signatures underlying variability in human memory consolidation. doi:10.1038/s41467-025-57766-x
44. Shin, G.-H., Kweon, Y.-S., Oh, S., et al. (2025). Personalized targeted memory reactivation enhances consolidation of challenging memories via slow wave and spindle dynamics. doi:10.1038/s41539-025-00340-3
45. Gibson, E. M., et al. (2014). Neuronal activity promotes oligodendrogenesis and adaptive myelination in the mammalian brain. doi:10.1126/science.1252304
46. McKenzie, I. A., et al. (2014). Motor skill learning requires active central myelination. doi:10.1126/science.1254960
47. Looser, Z. J., et al. (2024). Oligodendrocyte-axon metabolic coupling is mediated by extracellular K⁺ and maintains axonal health. doi:10.1038/s41593-023-01558-3
48. Hardingham, N. R., & Larkman, A. U. (1998). The reliability of excitatory synaptic transmission in slices of rat visual cortex in vitro is temperature dependent. doi:10.1111/j.1469-7793.1998.249bu.x
49. Volgushev, M., Vidyasagar, T. R., Chistiakova, M., Yousef, T., & Eysel, U. T. (2000). Membrane properties and spike generation in rat visual cortical cells during reversible cooling. doi:10.1111/j.1469-7793.2000.00059.x
50. Moser, E., Mathiesen, I., & Andersen, P. (1993). Association between brain temperature and dentate field potentials in exploring and swimming rats. doi:10.1126/science.8446900
51. Long, M. A., & Fee, M. S. (2008). Using temperature to analyse temporal dynamics in the songbird motor pathway. doi:10.1038/nature07448
52. Rangaraju, V., Calloway, N., & Ryan, T. A. (2014). Activity-driven local ATP synthesis is required for synaptic function. doi:10.1016/j.cell.2013.12.042
53. Rangaraju, V., Lauterbach, M., & Schuman, E. M. (2019). Spatially stable mitochondrial compartments fuel local translation during plasticity. doi:10.1016/j.cell.2018.12.013
54. Divakaruni, S. S., Van Dyke, A. M., Chandra, R., et al. (2018). Long-term potentiation requires a rapid burst of dendritic mitochondrial fission during induction. doi:10.1016/j.neuron.2018.09.025
55. Bapat, P., Nirschl, J. J., Wilkerson, J. R., et al. (2024). VAP stabilizes dendritic mitochondria to locally support synaptic plasticity. doi:10.1038/s41467-023-44233-8
56. Hu, H., Tang, J., Wu, Y., et al. (2025). Polarized ATP synthase in synaptic mitochondria induced by learning and plasticity signals. doi:10.1038/s42003-025-08963-3
57. Park, M., Salgado, J. M., Ostroff, L., Helton, T. D., Robinson, C. G., Harris, K. M., & Ehlers, M. D. (2006). Plasticity-induced growth of dendritic spines by exocytic trafficking from recycling endosomes. doi:10.1016/j.neuron.2006.09.040
58. Zhao, J., Fok, A. H. K., Fan, R., Kwan, P.-Y., Chan, H.-L., Lo, L. H.-Y., Chan, Y.-S., Yung, W.-H., Huang, J., Lai, C. S. W., & Lai, K.-O. (2020). Specific depletion of the motor protein KIF5B leads to deficits in dendritic transport, synaptic plasticity and memory. doi:10.7554/eLife.53456
59. Swarnkar, S., Avchalumov, Y., Espadas, I., Grinman, E., Liu, X.-A., Raveendra, B. L., Zucca, A., Mediouni, S., Sadhu, A., Valente, S., Page, D., Miller, K., & Puthanveettil, S. V. (2021). Molecular motor protein KIF5C mediates structural plasticity and long-term memory by constraining local translation. doi:10.1016/j.celrep.2021.109369
60. Rzechorzek, N. M., Thrippleton, M. J., Chappell, F. M., et al. (2022). A daily temperature rhythm in the human brain predicts survival after brain injury. doi:10.1093/brain/awab466
61. Ren, J., Sherry, A. D., & Malloy, C. R. (2015). 31P-MRS of healthy human brain: ATP synthesis, metabolite concentrations, pH, and T1 relaxation times. doi:10.1002/nbm.3384
62. Ren, J., Sherry, A. D., & Malloy, C. R. (2017). Efficient 31P band inversion transfer approach for measuring creatine kinase activity, ATP synthesis, and molecular dynamics in the human brain at 7 T. doi:10.1002/mrm.26560
63. Guo, R., Yang, S., Wiesner, H. M., et al. (2024). Mapping intracellular NAD content in entire human brain using phosphorus-31 MR spectroscopic imaging at 7 Tesla. doi:10.3389/fnins.2024.1389111
64. Kaiser, A., Vind, F. A., Duarte, J. M. N., et al. (2026). Ultra-high field 31P functional magnetic resonance spectroscopy reveals NAD+ dynamics in brain energy metabolism during visual stimulation. doi:10.1177/0271678X261415784
65. Karkouri, J., Novoselova, M., Rodgers, C. T., et al. (2026). Absolute Quantification of Brain Deuterium Metabolic Imaging in Healthy Volunteers and Glioblastoma Patients at 7T. doi:10.1002/mrm.70308
66. Li, X., Zhu, X.-H., Li, Y., et al. (2025). Quantitative mapping of key glucose metabolic rates in the human brain using dynamic deuterium magnetic resonance spectroscopic imaging. doi:10.1093/pnasnexus/pgaf072
67. Guo, R., Li, Y., Zhao, Y., et al. (2025). High-Resolution Brain Metabolic Imaging at Ultrahigh Field Using Extended Spatiospectral Encoding and Subspace Modeling. doi:10.1109/TBME.2025.3572448
68. Ahmadian, N., Karkouri, J., Deelchand, D. K., et al. (2025). Human Brain Deuterium Metabolic Imaging at 7 T: Impact of Different [6,6'-2H2]Glucose Doses. doi:10.1002/jmri.29532
69. Bøgh, N., Vaeggemose, M., Schulte, R. F., et al. (2024). Repeatability of deuterium metabolic imaging of healthy volunteers at 3 T. doi:10.1186/s41747-024-00426-4
70. Giese, K. P., Fedorov, N. B., Filipkowski, R. K., & Silva, A. J. (1998). Autophosphorylation at Thr286 of the alpha calcium-calmodulin kinase II in LTP and learning. doi:10.1126/science.279.5352.870
71. Lee, H.-K., Barbarosie, M., Kameyama, K., Bear, M. F., & Huganir, R. L. (2003). Regulation of distinct AMPA receptor phosphorylation sites during bidirectional synaptic plasticity. doi:10.1016/S0092-8674(03)00122-3
72. Rodrigues, S. M., Farb, C. R., Bauer, E. P., LeDoux, J. E., & Schafe, G. E. (2004). Pavlovian fear conditioning regulates Thr286 autophosphorylation of Ca2+/calmodulin-dependent protein kinase II at lateral amygdala synapses. doi:10.1523/JNEUROSCI.5303-03.2004
73. Tomita, S., Stein, V., Stocker, T. J., Nicoll, R. A., & Bredt, D. S. (2005). Bidirectional synaptic plasticity regulated by phosphorylation of stargazin-like TARPs. doi:10.1016/j.neuron.2005.01.009
74. Vierra, N. C., et al. (2023). Endoplasmic reticulum-plasma membrane junctions couple electrical activity to Ca2+-activated PKA signaling in neurons. doi:10.1038/s41467-023-40930-6
75. Biswas, D., et al. (2023). The landscape of the human brain phosphoproteome reveals region-specific phosphorylation events. doi:10.1021/acs.jproteome.2c00244
76. Shapson-Coe, A., Januszewski, M., Berger, D. R., et al. (2024). A petavoxel fragment of human cerebral cortex reconstructed at nanoscale resolution. doi:10.1126/science.adk4858
77. Johansen, A., Beliveau, V., Colliander, E., et al. (2024). An In Vivo High-Resolution Human Brain Atlas of Synaptic Density. doi:10.1523/JNEUROSCI.1750-23.2024
78. Lucchetti, F., Céléreau, E., Steullet, P., et al. (2025). Constructing the human brain metabolic connectome with MR spectroscopic imaging reveals cerebral biochemical organization. doi:10.1038/s41467-025-66124-w
79. van Blooijs, D., Nunes, A., van den Boom, M. A., et al. (2023). Developmental trajectory of transmission speed in the human brain. doi:10.1038/s41593-023-01272-0
80. Morgan, C. A., Thomas, D. L., Shao, X., et al. (2024). Measurement of blood-brain barrier water exchange rate using diffusion-prepared and multi-echo arterial spin labelling: Comparison of quantitative values and age dependence. doi:10.1002/nbm.5256
81. Baadsvik, E. L., Weiger, M., Froidevaux, R., Schildknecht, C. M., Ineichen, B. V., & Pruessmann, K. P. (2024). Myelin bilayer mapping in the human brain in vivo. doi:10.1002/mrm.29998
82. Chung, K. J., Abdelhafez, Y. G., Spencer, B. A., et al. (2025). Quantitative PET imaging and modeling of molecular blood-brain barrier permeability. doi:10.1038/s41467-025-58356-7
83. Farinas, A., Rutledge, J., Bot, V. A., et al. (2025). Disruption of the cerebrospinal fluid-plasma protein balance in cognitive impairment and aging. doi:10.1038/s41591-025-03831-3
84. Fultz, N. E., Bonmassar, G., Setsompop, K., et al. (2019). Coupled electrophysiological, hemodynamic, and cerebrospinal fluid oscillations in human sleep. doi:10.1126/science.aax5440
85. Kim, D., Huang, Y., & Liu, J. (2025). Non-invasive MRI measurements of age-dependent in vivo human glymphatic exchange using magnetization transfer spin labeling. doi:10.1016/j.neuroimage.2025.121142
86. Eide, P. K., Lashkarivand, A., Pripp, A., et al. (2023). Plasma neurodegeneration biomarker concentrations associate with glymphatic and meningeal lymphatic measures in neurological disorders. doi:10.1038/s41467-023-37685-5
87. Hirschler, L., Runderkamp, B. A. R., Decker, A., et al. (2025). Region-specific drivers of CSF mobility measured with MRI in humans. doi:10.1038/s41593-025-02073-3
88. Dagum, P., Elbert, D. L., Giovangrandi, L., et al. (2026). The glymphatic system clears amyloid beta and tau from brain to plasma in humans. doi:10.1038/s41467-026-68374-8
89. Finnema, S. J., Nabulsi, N. B., Mercier, J., et al. (2018). Kinetic evaluation and test-retest reproducibility of [¹¹C]UCB-J, a novel radioligand for positron emission tomography imaging of synaptic vesicle glycoprotein 2A in humans. doi:10.1177/0271678X17724947
90. Holiga, S., Sambataro, F., Luzy, C., et al. (2018). Test-retest reliability of task-based and resting-state blood oxygen level dependence and cerebral blood flow measures. doi:10.1371/journal.pone.0206583
91. Wirsich, J., Jorge, J., Iannotti, G. R., et al. (2021). The relationship between EEG and fMRI connectomes is reproducible across simultaneous EEG-fMRI studies from 1.5T to 7T. doi:10.1016/j.neuroimage.2021.117864
92. Vafaii, H., Mandino, F., Desrosiers-Grégoire, G., et al. (2024). Multimodal measures of spontaneous brain activity reveal both common and divergent patterns of cortical functional organization. doi:10.1038/s41467-023-44363-z
93. Amiri, M., Fisher, P. M., Raimondo, F., et al. (2023). Multimodal prediction of residual consciousness in the intensive care unit: the CONNECT-ME study. doi:10.1093/brain/awac335
94. Bolt, T., Wang, S., Nomi, J. S., et al. (2025). Autonomic physiological coupling of the global fMRI signal. doi:10.1038/s41593-025-01945-y
95. Manasova, V., et al. (2026). Multimodal multicentre investigation of diagnostic and prognostic markers in disorders of consciousness. doi:10.1093/brain/awaf412
96. Louveau, A., Smirnov, I., Keyes, T. J., et al. (2015). Structural and functional features of central nervous system lymphatic vessels. doi:10.1038/nature14432
97. Ahn, J. H., Cho, H., Kim, J.-H., et al. (2019). Meningeal lymphatic vessels at the skull base drain cerebrospinal fluid. doi:10.1038/s41586-019-1419-5
98. Kim, J., et al. (2025). Meningeal lymphatics-microglia axis regulates synaptic physiology. doi:10.1016/j.cell.2025.02.022
99. Eide, P. K., & Ringstad, G. (2021). Sleep deprivation impairs molecular clearance from the human brain. doi:10.1093/brain/awaa443
100. Qian, Y., Zhao, T., Zheng, H., Weimer, J., & Boada, F. E. (2012). High-resolution sodium imaging of human brain at 7 T. doi:10.1002/mrm.23225
101. Qian, Y., Lin, Y.-C., Chen, X., et al. (2025). Single-quantum sodium MRI at 3 T for separation of mono- and bi-T2 sodium signals. doi:10.1038/s41598-025-07800-1
102. Galarreta, M., & Hestrin, S. (1999). A network of fast-spiking cells in the neocortex connected by electrical synapses. doi:10.1038/47029
103. Anastassiou, C. A., Perin, R., Markram, H., & Koch, C. (2011). Ephaptic coupling of cortical neurons. doi:10.1038/nn.2727
104. Graydon, C. W., Cho, S., Diamond, J. S., Kachar, B., von Gersdorff, H., & Grimes, W. N. (2014). Specialized postsynaptic morphology enhances neurotransmitter dilution and high-frequency signaling at an auditory synapse. doi:10.1523/JNEUROSCI.4493-13.2014
105. Kilb, W., Dierkes, P. W., Syková, E., Vargová, L., & Luhmann, H. J. (2006). Hypoosmolar conditions reduce extracellular volume fraction and enhance epileptiform activity in the CA3 region of the immature rat hippocampus. doi:10.1002/jnr.20871
106. Xie, L., Kang, H., Xu, Q., Chen, M. J., Liao, Y., Thiyagarajan, M., O'Donnell, J., Christensen, D. J., Nicholson, C., Iliff, J. J., Takano, T., Deane, R., & Nedergaard, M. (2013). Sleep drives metabolite clearance from the adult brain. doi:10.1126/science.1241224
107. Lauderdale, K., Murphy, T., Tung, T., Davila, D., Binder, D. K., & Fiacco, T. A. (2015). Osmotic Edema Rapidly Increases Neuronal Excitability Through Activation of NMDA Receptor-Dependent Slow Inward Currents in Juvenile and Adult Hippocampus. doi:10.1177/1759091415605115
108. Burman, R. J., Brodersen, P. J. N., Raimondo, J. V., Sen, A., & Akerman, C. J. (2023). Active cortical networks promote shunting fast synaptic inhibition in vivo. doi:10.1016/j.neuron.2023.08.005
109. Yang, Y.-C., Wang, G.-H., Chou, P., Hsueh, S.-W., Lai, Y.-C., & Kuo, C.-C. (2024). Dynamic electrical synapses rewire brain networks for persistent oscillations and epileptogenesis. doi:10.1073/pnas.2313042121
110. Selfe, J. S., et al. (2024). All-optical reporting of inhibitory receptor driving force in the nervous system. doi:10.1038/s41467-024-53074-y
111. Voldsbekk, I., Maximov, I. I., Zak, N., Roelfs, D., Geier, O., Due-Tønnessen, P., Elvsåshagen, T., Strømstad, M., Bjørnerud, A., & Groote, I. (2020). Evidence for wakefulness-related changes to extracellular space in human brain white matter from diffusion-weighted MRI. doi:10.1016/j.neuroimage.2020.116682
112. Feld, G. B., Niethard, N., Liu, J., et al. (2026). Electrical Synapses Contribute to Sleep-Dependent Declarative Memory Retention. doi:10.1111/ejn.70401
113. Suzuki, A., et al. (2011). Astrocyte-neuron lactate transport is required for long-term memory formation. doi:10.1016/j.cell.2011.02.018
114. Silva, B., et al. (2022). Glial ketogenesis regulates memory maintenance during starvation. doi:10.1038/s42255-022-00528-6
115. Pavlowsky, A., et al. (2025). Neuronal fatty acid oxidation fuels memory after intensive learning in Drosophila. doi:10.1038/s42255-025-01416-5
116. Greda, A. K., et al. (2025). Interaction of sortilin with apolipoprotein E3 enables neurons to use long-chain fatty acids as alternative metabolic fuel. doi:10.1038/s42255-025-01389-5
117. Cahill, M. K., et al. (2024). Network-level encoding of local neurotransmitters in cortical astrocytes. doi:10.1038/s41586-024-07311-5
118. Williamson, N. R., et al. (2025). Learning-associated astrocyte ensembles regulate memory recall. doi:10.1038/s41586-024-08170-w
119. Dewa, K., et al. (2025). The astrocytic ensemble acts as a multiday trace to stabilize memory. doi:10.1038/s41586-025-09619-2
120. Villemagne, V. L., Harada, R., Dore, V., et al. (2022). First-in-Humans Evaluation of ¹⁸F-SMBT-1, a Novel ¹⁸F-Labeled Monoamine Oxidase-B PET Tracer for Imaging Reactive Astrogliosis. doi:10.2967/jnumed.121.263254
121. Villemagne, V. L., Harada, R., Dore, V., et al. (2022). Assessing Reactive Astrogliosis with ¹⁸F-SMBT-1 Across the Alzheimer Disease Spectrum. doi:10.2967/jnumed.121.263255
122. Hiraoka, K., Mesfin, B., Wu, Y., et al. (2025). Kinetic and quantitative analysis of [¹⁸F]SMBT-1 PET imaging for monoamine oxidase B. doi:10.1007/s12149-025-02083-y
123. Mesfin, B., Ishioka, Y., Ichinose, Y., et al. (2026). Whole-body biodistribution of [¹⁸F]SMBT-1: a novel PET tracer for monoamine oxidase B imaging in healthy humans. doi:10.1007/s12149-025-02144-2
124. Tyacke, R. J., Myers, J. F. M., Venkataraman, A., et al. (2018). Evaluation of ¹¹C-BU99008, a PET Ligand for the Imidazoline₂ Binding Site in Human Brain. doi:10.2967/jnumed.118.208009
125. Livingston, N. R., Calsolaro, V., Hinz, R., et al. (2022). Relationship between astrocyte reactivity, using novel ¹¹C-BU99008 PET, and glucose metabolism, grey matter volume and amyloid load in cognitively impaired individuals. doi:10.1038/s41380-021-01429-y
126. Best, L. M., Truong, J., McCluskey, T., et al. (2026). MAO-B status in alcohol use disorder: a [¹¹C]SL25.1188 PET imaging study of putative astrogliosis. doi:10.1038/s41380-025-03355-9
127. Jaisa-Aad, M., Muñoz-Castro, C., Healey, M. A., Hyman, B. T., & Serrano-Pozo, A. (2024). Characterization of monoamine oxidase-B (MAO-B) as a biomarker of reactive astrogliosis in Alzheimer's disease and related dementias. doi:10.1007/s00401-024-02712-2
128. Bukalo, O., et al. (2026). Astrocytes enable amygdala neural representations supporting memory. doi:10.1038/s41586-025-10068-0
129. Lee, J.-C., Wang, C.-Y., Lin, C.-L., & Lu, H.-C. (2022). Synaptic memory survives molecular turnover. doi:10.1073/pnas.2211572119
130. Musall, S., Kaufman, M. T., Juavinett, A. L., Gluf, S., & Churchland, A. K. (2019). Single-trial neural dynamics are dominated by richly varied movements. doi:10.1038/s41593-019-0502-4
131. Mostert, P., Albers, A. M., Brinkman, L., Todorova, L., & de Lange, F. P. (2018). Eye Movement-Related Confounds in Neural Decoding of Visual Working Memory Representations. doi:10.1523/ENEURO.0401-17.2018
132. McFarland, D. J., McCane, L. M., David, S. V., & Wolpaw, J. R. (2005). Brain-computer interface operation: signal and noise during early training sessions. doi:10.1088/1741-2560/2/4/014
133. Chaibub Neto, E., Pratap, A., Perumal, T. M., et al. (2019). Identity confounding in machine learning can be controlled by design. doi:10.1038/s41746-019-0178-x
134. Xu, M., Fanton, S., Jahanbekam, A., et al. (2020). The Cross-Dataset Variability Problem in EEG Decoding With Deep Learning. doi:10.3389/fnhum.2020.00103
135. Jiang, Y., Li, Y., Jia, Y., et al. (2024). Large Brain Model for Learning Generic Representations with Tremendous EEG Data in BCI. ICLR 2024 paper
136. Lee, N., Barmpas, K., Panagakis, Y., Adamos, D., Laskaris, N., & Zafeiriou, S. (2025). Are Large Brainwave Foundation Models Capable Yet? Insights from Fine-Tuning. PMLR 267
137. Han, D. D., Lee, A. L., Lee, T., et al. (2025). DIVER-0: A Fully Channel Equivariant EEG Foundation Model. arXiv:2507.14141
138. Chen, Z., Qin, C., You, W., et al. (2025). HEAR: An EEG Foundation Model with Heterogeneous Electrode Adaptive Representation. arXiv:2510.12515
139. El Ouahidi, Y., Lys, J., Thölke, P., et al. (2025). REVE: A Foundation Model for EEG -- Adapting to Any Setup with Large-Scale Pretraining on 25,000 Subjects. arXiv:2510.21585
140. Ma, J., Wu, F., Xing, Y., et al. (2026). Structured Prototype-Guided Adaptation for EEG Foundation Models. arXiv:2602.17251
141. Liu, D., Chen, Y., Chen, Z., et al. (2026). EEG Foundation Models: Progresses, Benchmarking, and Open Problems. arXiv:2601.17883
142. Lahiri, J. B., Runwal, P., Kulkarni, A., et al. (2026). PRISM: Exploring Heterogeneous Pretrained EEG Foundation Model Transfer to Clinical Differential Diagnosis. arXiv:2603.02268
143. EEG Challenge (2025). Official homepage. eeg2025.github.io
144. EEG Challenge (2025). Rules. Rules page
145. EEG Challenge (2025). Submission. Submission page
146. EEG Challenge (2025). Leaderboard. Leaderboard page
147. Murphy, K., Harris, A. D., & Wise, R. G. (2011). Robustly measuring vascular reactivity differences with breath-hold: normalising stimulus-evoked and resting state BOLD fMRI data. doi:10.1016/j.neuroimage.2010.07.059
148. Williams, R. J., Specht, J. L., Mazerolle, E. L., et al. (2023). Correspondence between BOLD fMRI task response and cerebrovascular reactivity across the cerebral cortex. doi:10.3389/fphys.2023.1167148
149. Yücel, M. A., Selb, J., Huppert, T. J., Franceschini, M. A., & Boas, D. A. (2015). Short separation regression improves statistical significance and better localizes the hemodynamic response obtained by near-infrared spectroscopy for tasks with differing autonomic responses. doi:10.1117/1.NPh.2.3.035005
150. Epp, S. M., Halani, S., Paquette, M., et al. (2025). BOLD signal changes can oppose oxygen metabolism across the human cortex. doi:10.1038/s41593-025-02132-9
151. Vinck, M., Oostenveld, R., van Wingerden, M., Battaglia, F., & Pennartz, C. M. A. (2011). An improved index of phase-synchronization for electrophysiological data in the presence of volume-conduction, noise and sample-size bias. doi:10.1016/j.neuroimage.2011.01.055
152. Haufe, S., Nikulin, V. V., Müller, K.-R., & Nolte, G. (2013). A critical assessment of connectivity measures for EEG data: A simulation study. doi:10.1016/j.neuroimage.2012.09.036
153. Palva, J. M., Wang, S. H., Palva, S., Zhigalov, A., Monto, S., Brookes, M. J., Schoffelen, J.-M., & Jerbi, K. (2018). Ghost interactions in MEG/EEG source space: A note of caution on inter-areal coupling measures. doi:10.1016/j.neuroimage.2018.02.032
154. Ye, S., Kitajo, K., & Kitano, K. (2020). Information-theoretic approach to detect directional information flow in EEG signals induced by TMS. doi:10.1016/j.neures.2019.09.003
155. Miljevic, A., Murphy, O. W., Fitzgerald, P. B., & Bailey, N. W. (2025). Estimating sensor-space EEG connectivity PART 1: Identifying best performing methods for functional connectivity in simulated data. doi:10.1016/j.clinph.2025.03.043
156. Bérut, A., Arakelyan, A., Petrosyan, A., et al. (2012). Experimental verification of Landauer’s principle linking information and thermodynamics. doi:10.1038/nature10872
157. Attwell, D., & Laughlin, S. B. (2001). An energy budget for signaling in the grey matter of the brain. doi:10.1097/00004647-200110000-00001
158. Lynn, C. W., Cornblath, E. J., Papadopoulos, L., et al. (2021). Broken detailed balance and entropy production in the human brain. doi:10.1073/pnas.2109889118
159. Deco, G., Sanz Perl, Y., Bocaccio, H., Tagliazucchi, E., & Kringelbach, M. L. (2022). The INSIDEOUT framework provides precise signatures of the balance of intrinsic and extrinsic dynamics in brain states. doi:10.1038/s42003-022-03505-7
160. de la Fuente, L. A., Zamberlan, F., Bocaccio, H., et al. (2023). Temporal irreversibility of neural dynamics as a signature of consciousness. doi:10.1093/cercor/bhac177
161. Nartallo-Kaluarachchi, R., Bonetti, L., Fernández-Rubio, G., et al. (2025). Multilevel irreversibility reveals higher-order organization of nonequilibrium interactions in human brain dynamics. doi:10.1073/pnas.2408791122
162. Ishihara, K., & Shimazaki, H. (2025). State-space kinetic Ising model reveals task-dependent entropy flow in sparsely active nonequilibrium neuronal dynamics. doi:10.1038/s41467-025-66669-w
163. Martínez, I. A., Bisker, G., Horowitz, J. M., & Parrondo, J. M. R. (2019). Inferring broken detailed balance in the absence of observable currents. doi:10.1038/s41467-019-11051-w
164. Blom, K., Song, K., Vouga, E., Godec, A., & Makarov, D. E. (2024). Milestoning estimators of dissipation in systems observed at a coarse resolution. doi:10.1073/pnas.2318333121
165. Grubb, M. S., & Burrone, J. (2010). Activity-dependent relocation of the axon initial segment fine-tunes neuronal excitability. doi:10.1038/nature09160
166. Thomas, C., Ye, F. Q., Irfanoglu, M. O., Modi, P., Saleem, K. S., Leopold, D. A., & Pierpaoli, C. (2014). Anatomical accuracy of brain connections derived from diffusion MRI tractography is inherently limited. doi:10.1073/pnas.1405672111
167. Reveley, C., Seth, A. K., Pierpaoli, C., Silva, A. C., Yu, D., Saunders, R. C., Leopold, D. A., & Ye, F. Q. (2015). Superficial white matter fiber systems impede detection of long-range cortical connections in diffusion MR tractography. doi:10.1073/pnas.1418198112
168. Donahue, C. J., Sotiropoulos, S. N., Jbabdi, S., Hernandez-Fernandez, M., Behrens, T. E., Dyrby, T. B., Coalson, T., Kennedy, H., Knoblauch, K., Van Essen, D. C., & Glasser, M. F. (2016). Using diffusion tractography to predict cortical connection strength and distance: A quantitative comparison with tracers in the monkey. doi:10.1523/JNEUROSCI.0493-16.2016
169. Schilling, K. G., Petit, L., Rheault, F., Remedios, S., Pierpaoli, C., Anderson, A. W., Landman, B. A., & Descoteaux, M. (2020). Brain connections derived from diffusion MRI tractography can be highly anatomically accurate if we know where white matter pathways start, where they end, and where they do not go. doi:10.1007/s00429-020-02129-z
170. Grisot, G., Haber, S. N., Hawrylycz, M., Yendiki, A., et al. (2021). Diffusion MRI and anatomic tracing in the same brain reveal common failure modes of tractography. doi:10.1016/j.neuroimage.2021.118300
171. Gajwani, M., Oldham, S., Pang, J. C., Arnatkevičiūtė, A., Tiego, J., Bellgrove, M. A., & Fornito, A. (2023). Can hubs of the human connectome be identified consistently with diffusion MRI? doi:10.1162/netn_a_00324
172. Sarwar, T., Ramamohanarao, K., Daducci, A., Schiavi, S., Smith, R. E., & Zalesky, A. (2023). Evaluation of tractogram filtering methods using human-like connectome phantoms. doi:10.1016/j.neuroimage.2023.120376
173. He, Y., Hong, Y., Wu, Y., et al. (2024). Spherical-deconvolution informed filtering of tractograms changes laterality of structural connectome. doi:10.1016/j.neuroimage.2024.120904
174. McMaster, E. M., Newlin, N. R., Rudravaram, G., et al. (2025). Harmonized connectome resampling for variance in voxel sizes. doi:10.1016/j.mri.2025.110424
175. Manzano-Patrón, J. P., Deistler, M., Schröder, C., et al. (2025). Uncertainty mapping and probabilistic tractography using Simulation-based Inference in diffusion MRI: A comparison with classical Bayes. doi:10.1016/j.media.2025.103580
176. Zhu, S., Huszar, I. N., Cottaar, M., et al. (2025). Imaging the structural connectome with hybrid MRI-microscopy tractography. doi:10.1016/j.media.2025.103498
177. Bramati, I. B., Szczupak, D., Carneiro Monteiro, M., Meireles, F., Menezes Guimarães, D., Dean, R. J., Paul, L. K., & Tovar-Moll, F. (2026). Diffusion MRI sampling schemes bias diffusion metrics and tractography. doi:10.3389/fnimg.2026.1670604
178. Huber, R., Mäki, H., Rosanova, M., Casarotto, S., Canali, P., Casali, A. G., Tononi, G., & Massimini, M. (2013). Human cortical excitability increases with time awake. doi:10.1093/cercor/bhs014
179. Kuhn, M., Wolf, E., Maier, J. G., Mainberger, F., Feige, B., Schmid, H., et al. (2016). Sleep recalibrates homeostatic and associative synaptic plasticity in the human cortex. doi:10.1038/ncomms12455
180. Zrenner, C., Desideri, D., Belardinelli, P., & Ziemann, U. (2018). Real-time EEG-defined excitability states determine efficacy of TMS-induced plasticity in human motor cortex. doi:10.1016/j.brs.2017.11.016
181. Fehér, K. D., Henckaerts, P., Hirsch, V., Bucsenez, U., Kuhn, M., Maier, J. G., et al. (2026). A nap can recalibrate homeostatic and associative synaptic plasticity in the human cortex. doi:10.1016/j.neuroimage.2026.121723
182. Santoni, G., et al. (2024). Chromatin plasticity predetermines neuronal eligibility for memory trace formation. doi:10.1126/science.adg9982
183. Wang, J., Telese, F., Tan, Y., et al. (2015). LSD1n is an H4K20 demethylase regulating memory formation via transcriptional elongation control. doi:10.1038/nn.4069
184. Dai, J., Aoto, J., & Südhof, T. C. (2019). Alternative splicing of presynaptic neurexins differentially controls postsynaptic NMDA and AMPA receptor responses. doi:10.1016/j.neuron.2019.03.032
185. Shi, H., Zhang, X., Weng, Y.-L., et al. (2018). m6A facilitates hippocampus-dependent learning and memory through YTHDF1. doi:10.1038/s41586-018-0666-1
186. Peterson, L. N., Kasper, J. M., Allgaier, J. A., et al. (2025). ADAR2-mediated Q/R editing of GluA2 in homeostatic synaptic plasticity. doi:10.1126/scisignal.adr1442
187. Joglekar, A., Prjibelski, A., Mahfouz, A., et al. (2024). Single-cell long-read sequencing-based mapping reveals specialized splicing patterns in developing and adult mouse and human brain. doi:10.1038/s41593-024-01616-4
188. Li, Y., Zhu, M., Li, X., et al. (2025). Enhanced Protein Synthesis and Hippocampus-Dependent Memory via Inhibition of YTHDF2-Mediated m6A mRNA Degradation. doi:10.1002/advs.202514926
189. Govindarajan, A., Israely, I., Huang, S.-Y., & Tonegawa, S. (2011). The dendritic branch is the preferred integrative unit for protein synthesis-dependent LTP. doi:10.1016/j.neuron.2010.12.008
190. Frischknecht, R., Heine, M., Perrais, D., Seidenbecher, C. I., Choquet, D., & Gundelfinger, E. D. (2009). Brain extracellular matrix affects AMPA receptor lateral mobility and short-term synaptic plasticity. doi:10.1038/nn.2338
191. Glykys, J., Dzhala, V., Egawa, K., et al. (2014). Local impermeant anions establish the neuronal chloride concentration. doi:10.1126/science.1245423
192. Seidl, A. H., Rubel, E. W., & Barria, A. (2015). Tuning of Ranvier node and internode properties in myelinated axons to adjust action potential timing. doi:10.1038/ncomms9073
193. Stringer, C., Pachitariu, M., Steinmetz, N., et al. (2019). Spontaneous behaviors drive multidimensional, brainwide activity. doi:10.1126/science.aav7893
194. Saleem, A. B., Ayaz, A., Jeffery, K. J., Harris, K. D., & Carandini, M. (2013). Integration of visual motion and locomotion in mouse visual cortex. doi:10.1038/nn.3567
195. Ravassard, P., Kees, A., Willers, B., et al. (2013). Multisensory control of hippocampal spatiotemporal selectivity. doi:10.1126/science.1232655
196. Zelano, C., Jiang, H., Zhou, G., et al. (2016). Nasal respiration entrains human limbic oscillations and modulates cognitive function. doi:10.1523/JNEUROSCI.2586-16.2016
197. Flesher, S. N., Downey, J. E., Weiss, J. M., et al. (2021). A brain-computer interface that evokes tactile sensations improves robotic arm control. doi:10.1126/science.abd0380
198. Raut, R. V., Rosenthal, Z. P., Wang, X., et al. (2025). Arousal as a universal embedding for spatiotemporal brain dynamics. doi:10.1038/s41586-025-09544-4
199. Guo, C., Pleiss, G., Sun, Y., & Weinberger, K. Q. (2017). On Calibration of Modern Neural Networks. PMLR 70:1321-1330
200. Geifman, Y., & El-Yaniv, R. (2017). Selective Classification for Deep Neural Networks. NeurIPS 2017
201. Ji, Z., et al. (2023). Survey of Hallucination in NLG. doi:10.1145/3571730
202. Correa, J. D., Lee, S., & Bareinboim, E. (2021). Nested Counterfactual Identification. arXiv:2107.03190
203. Gorgolewski, K. J., et al. (2016). BIDS. doi:10.1038/sdata.2016.44
204. Pernet, C. R., et al. (2019). EEG-BIDS. doi:10.1038/s41597-019-0104-8
205. Casali, A. G., et al. (2013). PCI. doi:10.1126/scitranslmed.3006294
206. Chalmers, D. J. (1995). Facing up to the problem of consciousness. Journal of Consciousness Studies, 2(3), 200-219.
207. Parfit, D. (1984). Reasons and Persons. Oxford University Press.