50-Worker Research Harvest: A Literature Map Organized by Unresolved Questions

Technical chronology policy for the 2025-2026 frontier

This evidence bank uses the core site's issue-year citation rule for the technical frontier, but raw year adjacency is not treated as one ladder. The reason is concrete. Nature lists Terceros et al. as Published: 26 November 2025 while the citation line is Nature volume 649, pages 1254-1263 (2026). Dewa et al. is Published: 15 October 2025 with issue date 04 December 2025. Hirschler et al. is Published: 14 October 2025, Dagum et al. is Published: 27 January 2026, and Bukalo et al. is Published: 11 February 2026. Therefore, U3 controller-side causality and U1/U7 bounded human observability must be named before year order is allowed to shape a frontier judgment.

Read tractography-derived human connectomes as route-conditioned macro pathway priors

One remaining compression at the literature front door was that connectome could still refer to a living-human diffusion-MRI tractography graph without naming which part of the pipeline created the claim. The primary literature does not support reading that object as one stable graph. Thomas et al. (2014) showed that tractography accuracy is inherently limited when only voxel-averaged local orientation estimates are available, Schilling et al. (2018) showed persistent gyral-endpoint bias across algorithms and diffusion models, and Schilling et al. (2020) showed that high anatomical accuracy appears mainly when strong start / end / exclusion priors are supplied, which is a targeted bundle-hypothesis route rather than generic edge-complete recovery. Downstream graph meaning is unstable too: Gajwani et al. (2023) showed across 1,760 group connectomes built from 40 pipelines and 44 group-reconstruction schemes that hub location depends strongly on processing choices, while McMaster et al. (2025) and Bramati et al. (2026) show that voxel-size harmonization and diffusion-sampling scheme still move tractography outputs. Even newer gains such as Zhu et al. (2025) improve reconstruction through hybrid MRI-microscopy calibration, not by turning living-human tractography into a finished connectome. Therefore, the safe U1/U7 reading is acquisition-, endpoint-, graph-construction-, and calibration-conditioned macro pathway prior or, at best, targeted bundle-hypothesis route, not synapse-resolved structural truth or a WBE-relevant wiring completion claim. For the operating rule, go next to Wiki: tractography route card and Verification: Observability Budget.

Read same-brain functional connectomics as scaffold, not as a solved local twin

One remaining compression at the literature front door was inside the same-brain lane itself. The primary literature does not support reading same-brain functional connectomics as one solved class. Bosch et al. (2022) showed that live physiology to EM correlation is a multistage landmark-based bridge, MICrONS Consortium et al. (2025) showed that same-brain structure-function linkage remains a sequential local pipeline, Ding et al. (2025) added a validated stimulus-conditioned response model, and Gamlin et al. (2025) still transferred transcriptomic identity through morphology-based prediction. At the remaining latent-state level, Molnár et al. (2016), Sakamoto et al. (2018), Holler et al. (2021), Dürst et al. (2022), Emperador-Melero et al. (2024), and Mittermaier et al. (2024) show that current synaptic efficacy, release-site number, active-zone nanostructure / priming-site assembly, release probability, and membrane-state-gated consolidation are not closed by structure-function correspondence, while Beiran & Litwin-Kumar (2025) show that connectome-constrained dynamics can remain degenerate until extra recordings narrow the solution space. Therefore, the safe U7/U8 reading is sequential scaffold plus task-bounded conditional prediction, not direct transcriptomic truth, presynaptic release-machinery truth, current synaptic-state readout, or one solved local twin.

Read U3 ECM / PNN papers as route-family claims, not as one support bucket

The literature front door still had one missing U3 split in the matrix lane. The primary literature does not support reading ECM / PNN evidence as one generic support variable. Pizzorusso et al. (2002) showed a plasticity-window reopening route, Frischknecht et al. (2009) showed a receptor-mobility / short-term-plasticity constraint route, Nguyen et al. (2020) showed a microglia-driven ECM-remodeling and remote-memory-precision route, Jabłońska et al. (2024) showed a synapse-specific inhibitory-plasticity route, Alexander et al. (2025) showed a CA2-versus-PV memory-support route, Mehak et al. (2025) showed an age-linked rescue route, and Lehner et al. (2024) plus Banovac et al. (2025) remain human ex vivo histology routes. Therefore, the safe U3 reading is route-family evidence about hidden maintenance variables, not a direct living-human whole-brain ECM meter or one solved support-state readout.

Read U3 phospho-signaling papers as route-family claims, not as one controller row

The literature front door still had one missing U3 split inside the molecular-maintenance lane. The primary literature does not support reading phospho-signaling / second-messenger evidence as one common controller. Lee et al. (2003) and Rodrigues et al. (2004) are phosphosite-specific plasticity and learning-linked local phosphorylation routes, Havekes et al. (2016) and Vierra et al. (2023) are compartmentalized second-messenger / signalosome routes, Barone et al. (2023) is a circadian phospho-timing gate, Altas et al. (2024) is a region-specific phosphorylation and synapse-localization route, Rodriguez et al. (2025) is a single-site phospho-mutant causal intervention, and Biswas et al. (2023) is a human ex vivo phosphoproteome atlas route. Therefore, the safe U3 reading is to name the route family first; these papers do not jointly provide a route-free current whole-brain phospho-controller, one universal memory controller, or a living-human in vivo phospho-state readout.

Read U3 human clearance papers as route-family claims, not as one transport bucket

The literature front door still had one remaining compression inside the human clearance lane. The current primary literature does not support reading human clearance-support evidence as one transport row. Fultz et al. (2019) is a macroscopic sleep-state CSF-oscillation route, Kim, Huang, & Liu (2025) is a parenchyma-CSF water-exchange route, Lim et al. (2025) is a respiration-conditioned net-flow route whose direct observable remains plane-specific awake-state CSF displacement, Yoo et al. (2025) is an exercise-conditioned contrast-influx / meningeal-lymphatic route, Eide et al. (2023) is an intrathecal-tracer / CSF-to-blood-clearance-capacity route, Hirschler et al. (2025) is a CSF-mobility MRI route, and Dagum et al. (2026) is a model-based overnight biomarker-efflux route. Therefore, the safe U3 reading is route-family evidence about bounded human transport-side observables, not route-free whole-brain bulk circulation, local immune-controller identity, or one solved human maintenance readout.

Read U3 blood-CSF-barrier papers as route-family claims, not as one BBB-adjacent support row

The literature front door still had one remaining compression inside the barrier-side human lane. The primary literature does not support reading blood-CSF barrier / choroid-plexus evidence as one reusable route next to BBB and clearance papers. Zhao et al. (2020) and Sun et al. (2024) are choroid-plexus perfusion routes, Petitclerc et al. (2021) is a blood-to-CSF water-transport route, Anderson et al. (2022) is a DCE water-cycling route, Wu et al. (2026) is an apparent BCSFB-exchange route, and Petitclerc et al. (2026) is a simultaneous BBB-versus-BCSFB ASL exchange route. Therefore, the safe U3 / U7 reading is route-family evidence about bounded blood-CSF-barrier observables, not generic BBB permeability, not route-free whole-brain clearance truth, and not direct choroid-plexus epithelial-controller identity.

Read U3 myelin / oligodendrocyte timing evidence as route-family claims, not as one delay row

The literature front door still had one remaining compression inside the timing-support lane. The primary literature does not support reading myelin / oligodendrocyte evidence as one common delay or one common human myelin row. Gibson et al. (2014) and McKenzie et al. (2014) are activity-dependent oligodendrogenesis / learning routes, Seidl et al. (2015) and Cohen et al. (2020) are node / internode / periaxonal timing-control routes, Xin et al. (2024) is a plasticity-brake route, and Della-Flora Nunes et al. (2025) is a recovery-boundary route showing that functional recovery can occur without complete remyelination. On the human side, van Blooijs et al. (2023) constrain a tract-scale transmission-speed route, Arshad et al. (2017) constrain an MWF versus calibrated T₁w/T₂w comparison route, Hagiwara et al. (2018) constrain a relaxometry / MT_sat comparison route, Baadsvik et al. (2024) constrain a bilayer-sensitive mapping route, Genc et al. (2025) constrain a developmental diffusion-microstructure route with ex vivo oligodendrocyte-expression alignment, Chen et al. (2025) show that orientation dependence remains an internal MT-family burden, Galbusera et al. (2025) constrain a qT1 remyelination-sensitive pathology route, and Colaes et al. (2026) show that T₁w/FLAIR may remain a general tissue-health marker rather than a myelin-specific readout. Therefore, the safe U3 reading is not myelin support exists; it has to name which mechanistic family or human proxy route is actually carrying the claim.

Read U8 closed-loop stability as more than fast timing

The U8 route also needed one more stop line. de Quervain et al. (1998) and Oei et al. (2007) showed that glucocorticoid state can impair retrieval and reduce human hippocampal / prefrontal retrieval activity. McCauley et al. (2020), Barone et al. (2023), and Birnie et al. (2023) showed that circadian and corticosteroid timing alter hippocampal plasticity machinery, while Reger et al. (2008) and Sherman et al. (2015) show that insulin signaling and circadian-rhythm consistency can shift human memory or hippocampal activity. Therefore, U8 papers are no longer read here from latency, jitter, decoder survival, or recalibration burden alone. The route now also asks whether slow internal-milieu variables such as circadian phase, glucocorticoid state, and insulin / metabolic regime were controlled, measured, perturbed, or left latent.

Why U4 now sits on its own rung

The site-wide route-card update already fixed DCM / effective-connectivity claims as model-conditioned causal hypotheses on the core pages, but this literature map still hid that rule inside a broader U4/U13 row. Penny et al. (2004), Rosa et al. (2012), Jafarian et al. (2020), Frässle et al. (2021), Jafarian et al. (2024), and Wu et al. (2024) together show why scaling and reliability do not erase candidate-model and observation-model dependence. Therefore, the technical route here now reads U4 before U13 so model-conditioned causal inference is not compressed into decode / imitation discussion.

Read U1 / U7 through the field-formation wall first

The remaining weakness in the literature route was that source validation and inverse-solver comparison still sat too close to the hidden assumption that the target source class had already reached the sensors. The primary literature does not support that shortcut. Ahlfors et al. (2010) showed strong orientation dependence in EEG / MEG sensitivity, Ahlfors et al. (2010) showed that extended or distributed sources can cancel at the surface, Goldenholz et al. (2009) showed that source extent and anatomy materially change cortical SNR, and Piastra et al. (2021) showed that EEG / MEG sensitivity depends on head-model detail including the CSF compartment. Therefore, on this page, U1 / U7 now asks technical readers to separate field-formation visibility, inverse uncertainty, and validation class rather than reading `validated ESI` as a general route to internal-state recovery.

Read wearable OPM-MEG as a field-control route, not as portable naturalistic readout

One more U1 / U7 compression still remained. The current primary literature does not support reading wearable OPM-MEG as if movement tolerance automatically removed the EEG / MEG visibility wall. Boto et al. (2018) established wearable feasibility but also exposed saturation risk without background-field control. Rea et al. (2021) and Mellor et al. (2022) show that precision field modeling and nulling remain part of the route, Holmes et al. (2025) show that lightly shielded operation still depends on active compensation plus tSSS, Rhodes et al. (2025) keep individual MRI as the gold standard even when pseudo-MRI is useful, Wu et al. (2025) show that crosstalk remains an array-level design burden, and Spedden et al. (2025) show whole-body stepping feasibility in only three healthy adults under a narrow sensorimotor beta task. Therefore, on this page a wearable OPM-MEG paper is first typed as a movement-tolerant acquisition route conditioned on shielding class, field nulling / interference suppression, calibration / coregistration, anatomy route, crosstalk burden, and task regime. It is not read as shield-free portability or as a separate escape from the EEG / MEG visibility / inverse wall.

Read U7 as more than synchronized clocks

The remaining weakness in the literature route was that U7 could still be read as a synchronization bucket only. The current primary literature does not support that shortcut. Kothe et al. (2025) show that LSL can support millisecond-scale synchronized acquisition for most neurobehavioral research, but still cannot infer device-side delay by itself. Vafaii et al. (2024) show that simultaneous Ca²⁺ and BOLD data contain both common and divergent network structure. Chen et al. (2025) show EEG-PET-MRI can recover coupled global dynamics together with distinct network patterns. Bolt et al. (2025) show that a major global fMRI mode is substantially coupled to autonomic physiology, and Epp et al. (2025) show that roughly 40% of significant task-related BOLD voxels can change opposite to oxygen metabolism. Amiri et al. (2023) further show that acute DoC EEG+fMRI same-sample models relied on a 48-patient subset, whereas Manasova et al. (2026) show that pairwise multimodal disagreements are higher in minimally conscious and improving patients. Therefore, on this page, U7 now asks technical readers to separate synchronization infrastructure, shared-vs-specific component evidence, quantity bridge / physiology grounding, and bundle robustness under missing-modality or cross-centre stress rather than reading multimodal as one monotonic ladder.

Read human measurement papers as their own evidence class

For technical reading, the first split is between destructive local structure and living-human in vivo proxy routes. Lu et al. (2023), Shapson-Coe et al. (2024), Dorkenwald et al. (2024), and MICrONS Consortium et al. (2025) belong to the destructive-route class because preservation, registration, throughput, and proofreading burden materially change what the structural result means. By contrast, Johansen et al. (2024), Lucchetti et al. (2025), Ren et al. (2015), Ren et al. (2017), Guo et al. (2024), Kaiser et al. (2026), Karkouri et al. (2026), Li et al. (2025), Baadsvik et al. (2024), Rzechorzek et al. (2022), Padrela et al. (2025), Chung et al. (2025), Zhao et al. (2020), Sun et al. (2024), Petitclerc et al. (2021), Anderson et al. (2022), Wu et al. (2026), Petitclerc et al. (2026), Fultz et al. (2019), Kim, Huang, & Liu (2025), Lim et al. (2025), Yoo et al. (2025), Eide et al. (2023), Hirschler et al. (2025), and Dagum et al. (2026) raise different living-human observability classes and still leave explicit latent-state ceilings. On this page, the in vivo papers are therefore read on three axes at once: what variable class the route constrains, how specialized or deployment-limited the route still is, and what bounded hidden-state family the route can safely calibrate. A regional SV2A atlas calibrates a synaptic-density comparison family, a five-metabolite ¹H-MRSI connectome calibrates a macro biochemical scaffold family, resting ³¹P balance calibrates a macro energetic-balance family, ³¹P MT calibrates a model-conditioned exchange-flux family, whole-brain ³¹P NAD mapping calibrates an intracellular NAD-content family, localized ³¹P fMRS calibrates a task-evoked NAD-dynamics family, deuterium absolute quantification calibrates an absolute metabolite-distribution family, dynamic deuterium MRSI calibrates a kinetic energetic-rate family, tract-scale transmission-speed or macro myelin / oligodendrocyte routes calibrate bounded timing-support proxy families, thermal papers calibrate a separate macro thermal-physiology family, BBB permeability and exchange papers calibrate a macro neurovascular / BBB proxy family, blood-CSF-barrier papers calibrate bounded choroid-plexus perfusion, blood-to-CSF transport, DCE water cycling, and apparent or simultaneous boundary-separated exchange families, and the human clearance papers split further into bounded sleep-state CSF-oscillation, parenchyma-CSF water-exchange, respiration-conditioned net-flow, exercise-conditioned contrast-influx / meningeal-lymphatic-flow, intrathecal-tracer / CSF-to-blood-clearance-capacity, CSF-mobility MRI, and model-based biomarker-efflux calibrators rather than one common clearance meter. Even inside the deuterium family, Ahmadian et al. (2025) show that dose changes downstream metabolite visibility and Bøgh et al. (2024) show that repeatability depends on the acquisition / time-point regime, so the family label still does not fix one route burden. The same no-compression rule now also applies inside the human intrinsic-excitability lane: Tallman et al. (2025) is a local clinical single-unit allocation-related readout, Huber et al. (2013), Kuhn et al. (2016), and Fehér et al. (2026) are sleep-history / plasticity-recalibration proxies, and Zrenner et al. (2018) plus Khatri et al. (2025) are state-gated perturbation proxies. Those human routes do not share one direct observable, and they still do not identify AIS geometry, ion-channel distribution, or the responsible homeostatic controller in vivo. None of those routes closes current post-transcriptional RNA-state, current phospho-signaling / second-messenger state, branch-local proteostasis / tag-capture, branch- or bouton-specific cargo-routing, chloride-set-point state, cell-specific neurovascular-controller state, or other cell-specific maintenance controllers. That is why this page now treats calibrator role as a separate evidence field rather than hiding it inside generic route maturity.

One more compression still remained inside the U7 spectroscopy lane. The primary literature does not support reading 1H-MRSI observability as one row. Lucchetti et al. (2025) built a five-metabolite parcel-similarity graph in living humans, whereas Guo et al. (2025) built a high-resolution metabolite-distribution route under explicit extended-spatiospectral-encoding and subspace-model burden. Those are not the same inferential object. On this page, `1H-MRSI` therefore never means one generic spectroscopy step: the safe reading has to separate macro biochemical similarity from high-resolution metabolite distribution before the route is compared with ³¹P or deuterium papers.

One more compression still remained inside the U7 human-observability lane. The primary literature does not support reading SV2A / synaptic-density PET as one solved row. Naganawa et al. (2024) constrained a tracer-specific quantification route for ¹⁸F-SynVesT-1, Johansen et al. (2024) built a healthy-human atlas, Matuskey et al. (2025) provided a disease contrast in autistic adults, Shatalina et al. (2024) linked [¹¹C]UCB-J to task switching and switch cost in healthy adults, Smart et al. (2021) showed that brief visual activation changes delivery but not binding, and Holmes et al. (2022) found no measurable overall SV2A change 24 h after ketamine despite symptom reduction. Those are different inferential slices. Therefore, on this page, SV2A PET now means only the named slice that the paper actually strengthens, not a direct readout of current synaptic efficacy or momentary synaptic state.

One more ceiling still had to stay explicit across the same-brain and U7 rows. The current primary literature does not support reading same-brain structure-function linkage or regional SV2A density as if they already fixed release-site number, docked-vesicle architecture, active-zone nanostructure / priming-site assembly, or current release competence. Molnár et al. (2016) showed multiple docked vesicles and multi-vesicular release in human synapses, Sakamoto et al. (2018) showed that Munc13-1 assemblies set independent release sites, Dürst et al. (2022) showed that vesicular release probability sets individual synaptic strength, and Emperador-Melero et al. (2024) showed that CaV2 clustering and vesicle priming are mediated by distinct active-zone machineries. Holler et al. (2021) and Mittermaier et al. (2024) strengthen same-brain scaffold and membrane-state gating, not a direct active-zone readout, while Smart et al. (2021) show that brief activation changes delivery without changing [¹¹C]UCB-J binding. Therefore, U7 uses same-brain scaffold and SV2A only as bounded calibrator classes unless a paper directly measures the presynaptic release machinery itself.

Anchor route	Direct observable	Safe calibrator role	Still not closed
Johansen et al. (2024)	Regional SV2A PET atlas in healthy humans	Healthy baseline for regional synaptic-density proxy comparisons	Current synaptic efficacy, release-site number, active-zone nanostructure / priming-site assembly, branch-local weights, tag state
Naganawa et al. (2024)	Tracer-specific noninvasive quantification of 18F-SynVesT-1 against a 1TC reference standard	Quantification-route calibration for an SV2A tracer family	Healthy atlas truth, disease contrast, release-site number, active-zone nanostructure / priming-site assembly, current release competence, task-linked state readout
Matuskey et al. (2025)	Case-control cortical [11C]UCB-J disease contrast in autistic adults	Disease-linked regional synaptic-density comparison slice	Universal baseline, momentary synaptic state, release-site number, current release competence, intervention-sensitive restoration truth
Shatalina et al. (2024)	[11C]UCB-J association with task-switching activity and switch cost in healthy adults	Task / cognition association slice for selected executive functions	All-task cognition meter, current synaptic efficacy, release-site number, active-zone nanostructure / priming-site assembly, momentary activation truth
Smart et al. (2021)	Brief visual activation changes tracer delivery without changing [11C]UCB-J binding	Activation-timescale ceiling for SV2A interpretation	Momentary release competence, release-site number, active-zone nanostructure / priming-site assembly, fast synaptic-efficacy readout
Holmes et al. (2022)	No measurable overall SV2A change 24 h after ketamine despite symptom reduction	Intervention-response ceiling for whole-brain SV2A interpretation	Rapid treatment-efficacy meter, release-site number, active-zone nanostructure / priming-site assembly, current release competence
Lucchetti et al. (2025)	Within-subject five-metabolite parcel-similarity graph	Macro biochemical scaffold / similarity calibration	Kinetic flux, cell-specific controller state, branch-local metabolism
Guo et al. (2025)	High-resolution whole-brain metabolite-distribution maps under extended spatiospectral encoding and subspace modeling	High-resolution metabolite-distribution calibration under an explicit acquisition / reconstruction burden	Parcel-similarity scaffold, kinetic flux, route-free reconstruction robustness, branch-local metabolism
Ren et al. (2015)	Resting 31P metabolite concentrations, ATP synthesis, and intra-/extracellular pH	Macro energetic-balance calibration in resting human brain	Exchange flux, localized NAD dynamics, branch-local ATP sufficiency
Ren et al. (2017)	31P magnetization-transfer PCr→ATP and Pi→ATP exchange-flux estimates	Model-conditioned macro exchange-flux calibration	Cell-specific ATP routing, localized controller identity, same-subject maintenance closure
Guo et al. (2024)	Whole-brain intracellular NAD-content map at 7 T	Macro 31P NAD-content calibration	Task-locked local NAD dynamics, cell-specific redox control, controller identity
Kaiser et al. (2026)	Functionally localized occipital-voxel 31P-fMRS NAD+ dynamics during visual stimulation	Localized task-evoked NAD-dynamics calibration	Whole-brain NAD map, resting energetic balance, branch-local mitochondrial state
Karkouri et al. (2026)	Absolute deuterated HDO / Glc / Glx / Lac maps under explicit quantification in a mixed healthy / glioblastoma 7 T workflow	Absolute deuterium metabolite-distribution calibration	Kinetic-rate identifiability without model / input assumptions, route-free dose / timing invariance, branch-local ATP reserve
Li et al. (2025)	Dynamic deuterium MRI / MRSI glucose transport and metabolic-rate maps under blood-input and kinetic modeling at a fixed 7 T operating point	Macro energetic-rate calibration under an explicit kinetic model	Branch-local ATP reserve, mitochondrial positioning, phospho state, route-free repeatability / portability
Ahmadian et al. (2025) / Bøgh et al. (2024)	Human deuterium-route dose sensitivity at 7 T and acquisition / time-point-conditioned repeatability at 3 T	Operating-point burden calibration for deuterium observability claims	Route-free dose invariance, field-independent repeatability, protocol portability, direct kinetic truth
van Blooijs et al. (2023) / Baadsvik et al. (2024) / Genc et al. (2025) / Galbusera et al. (2025) / Colaes et al. (2026)	Tract-scale transmission-speed estimates and quantity-defined macro myelin / oligodendrocyte-linked proxies	Bounded timing-support proxy calibration under named speed, contrast, bilayer, developmental-alignment, remyelination-pathology, or tissue-health-ceiling routes	Per-axon node / internode / periaxonal timing controller, recovery completeness, living-human timing-state ground truth
Rzechorzek et al. (2022)	Macro human thermal rhythm / temperature physiology	Bounded macro thermal-physiology calibration	Local operating temperature, device-heating confound, thermal-controller identity
Morgan et al. (2024) / Padrela et al. (2025) / Padrela et al. (2026)	Macro BBB water-exchange imaging under named ASL method families and cohort regimes	Bounded BBB water-exchange calibration under route-specific ASL fitting and interpretation	Cell-specific pericyte controller state, local BBB maintenance logic, amyloid-specific BBB truth, and branch-level support control
Chung et al. (2025)	Tracer-specific molecular BBB permeability imaging with dynamic PET and kinetic modeling	Bounded tracer-specific BBB transport calibration under named radiotracer and transport-model assumptions	Generic BBB permeability truth, cross-tracer equivalence, and cell-specific endothelial / pericyte controller state
Zhao et al. (2020) / Sun et al. (2024)	Human choroid-plexus perfusion imaging under ASL and lifespan cohort analysis	Bounded blood-CSF-barrier perfusion calibration	Blood-to-CSF transport truth, epithelial-transporter identity, and generic BBB equivalence
Petitclerc et al. (2021)	Human blood-to-CSF water transport under ultra-long-TE ASL	Bounded blood-CSF transport calibration	Choroid-plexus perfusion truth, DCE water cycling, route-free whole-brain clearance, and generic BBB equivalence
Anderson et al. (2022)	Human choroid-plexus water cycling with DCE-derived exchange terms	Bounded choroid-plexus water-cycling calibration	Perfusion equivalence, blood-to-CSF transport equivalence, and direct epithelial-controller identity
Wu et al. (2026) / Petitclerc et al. (2026)	Apparent BCSFB exchange and simultaneous BBB-versus-BCSFB exchange separation	Bounded boundary-separated exchange calibration under REXI or multi-TE ASL compartment models	One generic barrier scalar, route-free transport equivalence, and direct choroid-plexus epithelial-controller truth
Tallman et al. (2025)	Hippocampal single-unit sparse-code / firing-rate relation during episodic encoding and retrieval in epilepsy patients	Local clinical-unit allocation-related readout for human episodic-memory excitability	Pre-existing intrinsic excitability, AIS / channel state, synaptic-drive contribution, whole-brain controller coverage
Huber et al. (2013) / Kuhn et al. (2016) / Fehér et al. (2026)	Wake / sleep / nap dependent shifts in TMS-EEG excitability or PAS-induction plasticity	Bounded sleep-homeostasis / plasticity-recalibration proxy	Cell-specific controller identity, AIS state, synapse-specific mechanism, whole-brain excitability map
Zrenner et al. (2018) / Khatri et al. (2025)	EEG-defined or personalized whole-brain state conditioned differences in TMS-induced plasticity or corticospinal responses	Bounded state-gated perturbation proxy for human excitability-dependent response	AIS geometry, ion-channel distribution, allocation state, long-horizon homeostatic controller
Fultz et al. (2019)	Macroscopic CSF oscillations during human NREM sleep	Bounded sleep-state CSF-oscillation calibration	Net molecular clearance flux, protein-efflux truth, cell-specific immune controller state
Kim, Huang, & Liu (2025)	Parenchyma-CSF water exchange measured with MT spin labeling	Bounded parenchyma-CSF water-exchange calibration	Protein-clearance capacity, local immune controller identity, synapse-resolved maintenance control
Lim et al. (2025)	Plane-specific awake-state CSF displacement and net-flow changes linked to respiration and diaphragm motion	Bounded respiration-conditioned net-flow calibration	Route-free whole-brain bulk circulation, protein-clearance capacity, local immune controller identity
Yoo et al. (2025)	Intravenous-contrast-derived putative glymphatic influx and parasagittal meningeal-lymphatic flow after long-term exercise	Bounded intervention-conditioned contrast-influx / meningeal-lymphatic calibration	Natural-sleep whole-brain clearance truth, route-free local immune control, synapse-resolved maintenance control
Eide et al. (2023)	Intrathecal gadobutrol retention and pharmacokinetic CSF-to-blood clearance variables	Bounded intrathecal-tracer / CSF-to-blood-clearance-capacity calibration	Natural-sleep whole-brain clearance truth, local drainage-segment assignment, cell-specific immune controller state
Hirschler et al. (2025)	Region-specific CSF-mobility MRI and driver mapping	Bounded CSF-mobility calibration	Direct flux ground truth, protein-efflux truth, synapse-resolved clearance control
Dagum et al. (2026)	Model-based overnight Aβ / tau efflux to plasma in a randomized crossover sleep study	Bounded model-based biomarker-efflux calibration	Direct segmental drainage assignment, local immune-controller identity, moment-to-moment maintenance state

Read same-subject / same-brain papers as bridge-limited evidence

The next overread to stop is to treat same-subject or same-brain as if those labels already solved same-state continuity. The primary literature does not support that shortcut. Lu et al. (2023) show that preservation route changes extracellular-space retention and downstream ultrastructure, Bosch et al. (2022) show that correlative live-to-EM work is a landmark-based multistage bridge, MICrONS Consortium et al. (2025) show that same-brain function plus EM remains a sequential local pipeline, and Egger et al. (2024) show that even repeated live EEG can drift over a 10-hour window enough to motivate adaptive decoders. Therefore, on this page, specimen identity is read only as one bridge ingredient, not as same-state evidence by default.

Read bridge risk as family-specific, not as one generic time penalty

The next correction is narrower. The primary literature does not support treating bridge burden as one scalar that grows only with elapsed time. Lu et al. (2023) and Idziak et al. (2023) show that live-to-fix bridges are already transformation-dominated. Musall et al. (2019), Benisty et al. (2024), and Egger et al. (2024) show that same-day repeated live measurements can drift through spontaneous behavior, connectivity structure, and decoder-relevant EEG dynamics within hours. Hengen et al. (2016) and Xu et al. (2024) then show that sleep / wake crossing changes homeostatic and computational regime rather than merely adding more delay. Therefore, on this page, readers should ask not only how long the bridge was but also which hidden-state families were most exposed by that bridge type. The operating matrix is summarized in Wiki: State-Continuity Bridge and enforced in Verification: State-Continuity Bridge Card.

Read neuromodulatory papers as a route-family split

The literature route for U3 also needed one more correction. The current primary literature does not support reading all neuromodulatory papers as one human-state meter. Reimer et al. (2016) is a mixed arousal proxy, Lohani et al. (2022) and Neyhart et al. (2024) are local acetylcholine sensing routes, Hansen et al. (2022) and Goulas et al. (2021) are receptor / transporter atlas priors, Wong et al. (2013) is occupancy PET, and Koepp et al. (1998) plus Erritzoe et al. (2020) are challenge-linked displacement / release-sensitive PET. Therefore, on this page, U3 now asks readers to name the inferential object, time window, spatial scope, and model / challenge burden before any neuromodulatory paper is read as evidence about the current whole-brain transmitter state.

Read shared extracellular / electrical-state papers as a separate U3 route

Another missing U3 split was still hiding inside generic support language. Galarreta & Hestrin (1999) is a gap-junction coupling-network route, Anastassiou et al. (2011) is an endogenous-field / ephaptic route, Graydon et al. (2014) is a local extracellular-space geometry / transmitter-dilution route, Kilb et al. (2006) plus Xie et al. (2013) show osmotic or sleep-linked extracellular-space shifts, Voldsbekk et al. (2020) remains a bounded human diffusion-MRI proxy clue, and Feld et al. (2026) remains a human perturbation-conditioned electrical-synapse clue with pharmacological caveats. Therefore, on this page, U3 no longer lets shared extracellular / electrical state act as one common hidden-state meter. Readers now have to name whether the paper advances coupling, field effects, extracellular geometry / diffusion barriers, osmotic or sleep-linked regime shift, or only a bounded human clue.

Read maintenance-state papers as more than support background

Recent maintenance-state papers also changed in kind, not only in number. Hadzibegovic et al. (2025), Benoit et al. (2025), Hengen et al. (2016), Tallman et al. (2025), Huber et al. (2013), Kuhn et al. (2016), Zrenner et al. (2018), Khatri et al. (2025), Fehér et al. (2026), Alfonsa et al. (2025), Bell et al. (2010), Pandey et al. (2023), Mai-Morente et al. (2025), Terceros et al. (2026), Peterson et al. (2025), Vierra et al. (2023), Pandey et al. (2021), Aiken & Holzbaur (2024), Vishwanath et al. (2026), Kim et al. (2025), Suzuki et al. (2011), Silva et al. (2022), Pavlowsky et al. (2025), Greda et al. (2025), Dewa et al. (2025), and Bukalo et al. (2026) do not merely say that "support variables matter." They sharpen which hidden-state families remain outside connectome-only reading: intrinsic-excitability allocation / early engram bias, AIS / channel-state plasticity, homeostatic set-point / recovery control, human local clinical-unit allocation, human sleep-homeostasis / plasticity proxies, human state-gated perturbation proxies, ionic / chloride regulation, transcriptional stabilization gates, post-transcriptional RNA control, phospho-signaling / second-messenger routing, local proteostasis / tag-capture balance, cargo-routing state, bioenergetic / mitochondrial support, glial substrate-routing across lactate, ketone-body, fatty-acid, and apoE / sortilin-linked lipid routes, neurovascular-unit / BBB / pericyte support, clearance / immune support, astrocyte multiday traces, and astrocyte-enabled neural representations. The human intrinsic-excitability papers are therefore not one common meter but a bounded subfamily split with different ceilings and different missing controllers.

Read U3 sleep replay papers as route-family claims, not as one overnight-memory row

One more U3 split was still missing from the literature route. The current primary literature does not support reading sleep replay, TMR, and closed-loop sleep stimulation as if they already formed one reusable maintenance meter. Ngo et al. (2013) is a phase-locked slow-oscillation stimulation route, Whitmore et al. (2022) shows that benefit depends on ample and undisturbed slow-wave sleep, Baxter et al. (2023) shows that oscillation gains can occur without extra motor-memory gain, Geva-Sagiv et al. (2023) is an intracranial hippocampal-prefrontal synchrony intervention route, Schreiner et al. (2024) is a spindle-locked ripple route, Whitmore et al. (2024) shows memory-age dependence under sleep disruption, Jourde et al. (2025) shows that spindle-targeted auditory stimulation can either amplify sigma or truncate the spindle depending on timing, Duan et al. (2025) shows item-level electrophysiological variability in human consolidation, Deng et al. (2025) shows a time-windowed NREM physiology gate, and Shin et al. (2025) shows a difficulty-conditioned personalized TMR route. Therefore, on this page, U3 now requires technical readers to name the cueing or intervention route, sleep-integrity / disturbance burden, NREM physiology gate, oscillation-versus-memory effect split, and memory-selection / age regime before any sleep replay paper is read as evidence about a maintenance controller.

Read glial substrate-routing as a separate U3 route

One remaining U3 compression was still hiding between neuronal bioenergetic support and macro human energetic proxy progress. The current primary literature does not support reading those papers as one metabolic-support lane. Suzuki et al. (2011) constrain a lactate-shuttle support route, Silva et al. (2022) constrain a glia-to-neuron ketone-body route under starvation, Pavlowsky et al. (2025) constrain a glia-to-neuron fatty-acid route during intensive learning, and Greda et al. (2025) constrain apoE / sortilin-dependent neuronal lipid uptake and fuel-choice gating. By contrast, the living-human 31P, deuterium, and astrocyte-related PET rows on this site remain macro energetic or target-defined proxy classes, not direct identification of the active glial supplier, fuel class, transport route, or neuronal sink. Therefore, the safe U3 reading now has to say whether the paper advances local neuronal bioenergetic control, a named glial substrate-routing route family, or only a bounded human proxy class.

Read U3 post-transcriptional RNA papers as route-family claims, not as one RNA row

The literature route still had one remaining compression inside U3. The current primary literature does not support reading post-transcriptional RNA evidence as one generic maintenance controller or one reusable m6A slot. Wang et al. (2015) is a neuron-specific splice-isoform route whose downstream object is chromatin / transcriptional elongation control. Dai et al. (2019) is a splice-dependent transsynaptic receptor-balance route. Shi et al. (2018) is a YTHDF1 translation route, whereas Zhuang et al. (2023) and Li et al. (2025) are YTHDF2-mediated decay / stability routes with different scope: dentate-gyrus-specific reader assignment plus mossy-fiber / MF-CA3 development on the one hand, and forebrain-scale enhancement of activity-dependent protein synthesis and hippocampus-dependent memory on the other. Peterson et al. (2025) is an ADAR2 / GluA2 editing route for homeostatic scaling, and Joglekar et al. (2024) is a long-read atlas / observability-ceiling route rather than a living-human whole-brain in vivo measurement route. Therefore, on this page, U3 no longer lets one RNA paper stand in for current whole-brain RNA-controller state, one universal hippocampal m6A-reader assignment, or one solved human observability class. Readers now have to name whether the paper advances splice-isoform control, transsynaptic splice-dependent receptor balance, m6A translation, dentate-gyrus- or forebrain-scale YTHDF2-mediated decay / stability, RNA editing, or atlas ceiling.

Read U3 local proteostasis papers as a route-family split

The literature route also still had one remaining compression inside U3. The current primary literature does not support reading local proteostasis evidence as one generic consolidation controller. Frey & Morris (1997) and Shires et al. (2012) are tag / capture eligibility routes. Govindarajan et al. (2011) is a branch-level integration route. Fonseca et al. (2006), Pandey et al. (2021), and Chalatsi et al. (2026) are synthesis-degradation / autophagy-linked proteostasis routes, but at different controller scales: late-LTP maintenance balance, translation-coupled long-term-memory formation, and PVALB-interneuron proteostasis / excitability with hippocampus-dependent memory, respectively. Lee et al. (2022) and Thomas et al. (2025) are turnover-resistant persistence or candidate tag-substrate routes. Therefore, on this page, U3 no longer lets one proteostasis paper stand in for current whole-branch capture readiness, one generic autophagy controller, or one solved human observability class. Readers now have to name whether the paper advances tag / capture eligibility, branch-level integration, synthesis-degradation balance, autophagy-linked remodeling, inhibitory-circuit proteostasis / excitability, or turnover-resistant persistence / candidate tag substrate.

Read U3 cargo-routing papers as a route-family split

The literature route still had one more maintenance-side compression. The current primary literature does not support reading cargo-routing evidence as one generic trafficking background. Park et al. (2006) and Correia et al. (2008) are postsynaptic AMPAR / recycling-endosome delivery routes. Uchida et al. (2014) and Wong et al. (2024) are transport-path gating or local vesicle-confinement routes. Nakayama et al. (2017), Liau et al. (2023), and Espadas et al. (2024) are dendritic / synaptic RNA-cargo organization routes. de Queiroz et al. (2025) is an axonal RNA-localization route required for long-term memory, and Aiken & Holzbaur (2024) is a presynaptic cargo-delivery / pausing route in human neurons. Therefore, on this page, U3 no longer lets one cargo paper stand in for whole-neuron delivery correctness, one generic RNA-transport controller, or proof that the right receptors, RNA cargoes, or presynaptic components reached the right branch, spine, or bouton. Readers now have to name whether the paper advances postsynaptic receptor delivery, transport-path gating / local vesicle confinement, dendritic or synaptic RNA-cargo organization, axonal RNA localization, or presynaptic cargo retention / pausing.

Do not let chronology fuse U3 causality with U1/U7 observability

One more correction is still necessary. U3 maintenance-state causality and U1/U7 living-human observability do not move on one common ladder just because both accelerated in 2025-2026. Hadzibegovic et al. (2025), Terceros et al. (2026), Dewa et al. (2025), and Bukalo et al. (2026) sharpen controller-side or local-circuit causal dependence, whereas Lucchetti et al. (2025), Hirschler et al. (2025), and Dagum et al. (2026) sharpen bounded human observability classes. These papers differ in species, spatial unit, and inferential object. The date rule matters in practice: Nature lists Terceros et al. as Published: 26 November 2025 while the citation line is Nature volume 649, pages 1254-1263 (2026); Dewa et al. is Published: 15 October 2025 with issue date 04 December 2025; Hirschler et al. is Published: 14 October 2025; Dagum et al. is Published: 27 January 2026; and Bukalo et al. is Published: 11 February 2026. If this evidence bank mixes those conventions or lets adjacent dates stand in for one ladder, chronology itself begins to collapse two ladders into one. Therefore, on this page, technical readers should name whether a paper advances local causal relevance or bounded human observability before letting the year ordering influence the frontier judgment.

Read neurovascular / BBB evidence as a separate U3 route

The next missing split was inside U3 itself. Bell et al. (2010), Kisler et al. (2020), Pandey et al. (2023), Swissa et al. (2024), and Mai-Morente et al. (2025) do not describe a generic vascular nuisance. They split into pericyte controller causality, neurovascular coupling dependence, activity-linked BBB modulation, and capillary support for memory-related function. Meanwhile, Padrela et al. (2025) and Chung et al. (2025) are important because they raise a human BBB permeability / exchange proxy route, but that route still remains a macro proxy rather than a direct readout of cell-specific neurovascular maintenance control. Therefore, on this page, U3 now keeps controller-side neurovascular biology separate from human BBB proxy progress instead of hiding both inside clearance or generic support language.

Read U10 as a route-family split, not as one thermodynamic frontier

On the core pages, thermodynamic claims already require a route card, but this evidence bank had still left U10 too close to a generic "physical grounding" bucket. Lynn et al. (2021) estimated coarse-grained entropy-production lower bounds from fMRI state transitions, de la Fuente et al. (2023) used inversion decoding on ECoG, Nartallo-Kaluarachchi et al. (2025) used multilevel visibility-graph irreversibility on MEG, Ishihara & Shimazaki (2025) estimated model-based entropy flow from spike ensembles, and Epp et al. (2025) showed that BOLD changes can oppose oxygen-metabolism changes. Therefore, on this page, U10 is now read as a split among route families plus a separate physiology-side grounding burden, not as one common thermodynamic measurement frontier.

If you want paper-level anchors before the U map

The default route on this page is still organized by unresolved question. If you already know that you want only the 2025-2026 technology / natural-science frontier papers first, jump directly to the Paper Collection: 2025-2026 technical-only shortlist. That table gives concrete anchor papers for EEG foundation-model governance, field-formation visibility, destructive ultrastructure audit, living-human observability, state-continuity bridge limits, neurovascular / BBB route-family splits, shared extracellular / electrical-state route-family splits, neuromodulatory route-family splits, sleep replay / replay-coupling route-family splits, proteostasis / cargo route-family splits, maintenance-state boundary papers, direct source validation, closed-loop communication, and thermodynamic route-family splits before you return here to place them in U1/U7/U4/U13/U8/U3/U10.

Groups kept off the main route this round

U0 / U12 / U15 still matter, but they are not placed on the default technical route. The reason is that what needs to be fixed first here is not metaphysics or legal theory, but what is currently measurable, how far direct validation goes, where the closed loop breaks, and which hidden states remain. These groups are easier to read correctly after the experimental front has already been mapped.

Do not confuse status labels with strength of evidence

source_logged means intake accepted, curated means organized, and noise_excluded means intentionally excluded. These are separate from venue labels such as Scopus or arXiv, separate from literature-type labels such as Review or Media, and separate from evidence class. If those axes start to blur together, go back to the Wiki: Source Types, Status Labels, and Evidence Classes.

Operational note

This section is the "input acceptance log". The acceptance/rejection decision (U map reflection, citation priority, noise exclusion) will be determined in subsequent regular updates according to the quality gate procedure.

This is not the default entry point for technical and natural-science readers

This section is an intake queue, not a frontier-ranking page. If you want to follow the primary evidence in technology and the natural sciences, start instead from the Technology and Natural Sciences priority route above, or from U1 / U7 first, then the state-continuity bridge between U7 and U8, and then U4 / U13 / U8 / U3 / U10 in the main text.

How to read this table

Here, ID is the name tag of the problem, Current status is how far it has been solved, and Number of citations is the amount of relevant evidence. A large number of citations does not necessarily mean that the conclusion is fixed.

Default reading order for this table

For readers coming from technology and the natural sciences, the default route is U1 / U7 → bridge (U7 / U8) → U4 → U13 → U8 → U3 → U10. U11 is treated as a secondary route for experimental comparison, while U0 / U12 / U15 are treated as auxiliary routes that branch off and then return to the main line.

Why you do not need to be intimidated by the numbering

The U number is an internal management code and is not meant to be memorized. It is enough to first find a cluster that is close to you in the "big problem group" above, and then go down to the individual U's.

How not to read the status labels as a strength ranking

A partial solution indicates that some foundation is present, and an exploration stage indicates that the comparative basis is still weak. Unstandardized, Insufficient, and Undeveloped differ in the type of what is missing. If you want to see the difference in one page, please refer to Wiki: How to read partial solution/exploration stage/undeveloped.

Why this provenance correction matters

This evidence bank does not let a review article inherit the status of a controller-side primary result, and it does not let a news URL remain visible as if it were an evidence class once the primary paper is known. Review papers can map route families, and secondary coverage can flag a lead, but integrated evidence on this page is keyed to the verified source class named in the cited row.

Deduplication rules applied this round

• Separate the evaluation axes of "measurement," "causation," and "operation" and create separate questions even on the same theme.
• Literature prioritizes primary research, and news articles are retained as supplementary references with primary research links.
• Replaced documents with low relevance/duplication tendency, and maintained the number of documents for each U.

Organization policy for this round

The information for Rounds 1 to 114 was not deleted, but the roles of the public text and automation/ were re-separated. The reader will be shown a summary and judgment materials in the main text, and machine processing results and operation logs will be saved in CSV/audit memo.

2026-03-27 addendum: U1 is three coupled subroutes, not one progress bar

The weak point of this U1 summary was to place probabilistic inverse solvers, conductivity-sensitive forward models, and direct-validation benchmarks under one shared "inverse progress" label. The primary literature does not support that shortcut. Luria et al. (2024), Tong et al. (2025), and Feng et al. (2025) improve how candidate sets, uncertainty maps, or debiased inference can be exposed inside a stated inverse family. Rimpiläinen et al. (2019), Vorwerk et al. (2024), Vorwerk et al. (2025), and Vorwerk et al. (2026) show that conductivity modelling and estimation materially move EEG and even MEG source results. Mikulan et al. (2020), Pascarella et al. (2023), Unnwongse et al. (2023), and Hao et al. (2025) provide focal-source and clinical validation routes, but they do not define one universal board for focal, sparse, extended, and spontaneous regimes. Therefore this page now reads U1 through three coupled questions: how the candidate set is represented, how forward-model uncertainty is propagated, and which validation class / source regime was actually tested.

2026-03-30 deepening: why U3 now has to be read as route families, not one support bucket

The weak point here was not lack of volume but lack of inferential separation. This evidence bank still mixed generic PubMed placeholders, reviews, and controller-side primary papers in a way that let astrocyte-state, pericyte / BBB control, clearance / immune support, and human support-state proxy routes sound closer than they are. The primary literature does not support that compression. Dewa et al. (2025) is a multiday astrocytic memory-stabilization route, Bukalo et al. (2026) is an astrocyte-enabled amygdala representation route, and Xin et al. (2025) is a stress-linked neuron-astrocyte coupling route. On the neurovascular side, Kisler et al. (2020), Pandey et al. (2023), Swissa et al. (2024), and Mai-Morente et al. (2025) constrain different controller objects rather than one common vascular scalar. On the clearance side, Kim et al. (2025) is a meningeal-lymphatics / microglia synaptic-physiology route, whereas Padrela et al. (2025), Chung et al. (2025), Hirschler et al. (2025), and Dagum et al. (2026) are bounded human macro proxy routes with different direct observables and model burdens. Therefore, U3 is now written so technical readers must name the route family, direct observable, and human ceiling before talking about a minimum biological subject model.

2026-03-19 deepening: why U4 now needs route-card reading inside the literature map

The weak point here was not that the site already separated DCM from SCM on theory pages. The weak point was that this evidence bank still allowed whole-brain scale, faster inversion, or repeatability under matched conditions to sound closer than they are to solved identifiability. Penny et al. (2004) fixed that DCM inference is relative to the models being compared, Rosa et al. (2012) showed that very large model spaces can be searched efficiently from one full model, Jafarian et al. (2020) showed that observation-model choices such as neurovascular coupling assumptions remain part of the inference target, Frässle et al. (2021) pushed directed-connectivity estimation to whole-brain human fMRI, Jafarian et al. (2024) showed that reliability can be strong under closely matched resting-state MEG conditions, and Wu et al. (2024) reduced computational burden further. Therefore, on this site, these papers raise tractability, disclosed assumptions, and reliability under named conditions, but do not by themselves raise U4 to discovered causal wiring or unique mechanism recovery.

When stopped at the entrance of U10

This section is a bibliographic map, so the descriptions of Landauer, NESS, and EPR are condensed. If you want to clarify only the beginning of the meaning, it will be easier to follow if you look at Wiki: Thermodynamic grounding basics first.

2026-03-20 addendum: U10 must separate route families before reading "thermodynamic consistency"

The remaining weakness in U10 was that it still read too much like one bucket about energy cost. The primary literature does not support that compression. Lynn et al. (2021) estimated entropy-production lower bounds only after coarse-graining fMRI dynamics into macrostates, de la Fuente et al. (2023) used inversion decoding on ECoG and showed dependence on feature choice and model complexity, Nartallo-Kaluarachchi et al. (2025) measured multilevel irreversibility from MEG visibility-graph structure, and Ishihara & Shimazaki (2025) estimated model-based entropy flow from spiking populations under explicit kinetic-Ising assumptions. Meanwhile, Epp et al. (2025) showed that BOLD changes can oppose oxygen-metabolism changes, which means energetic wording still needs physiology-side grounding rather than brain-signal irreversibility alone. On this page, U10 now separates brain-signal route family, coarse-graining / state definition, model burden, and physiology-side grounding.

2026-04-03 addendum: U10 also needs reverse-transition, memory-order, and stability audits

Route-family disclosure alone was still not enough. The current primary literature does not support reading a clean irreversibility estimate as operationally comparable by default. Martínez et al. (2019) showed that waiting-time asymmetry can reveal hidden dissipation even when observable current vanishes, Hartich & Godec (2024) showed that this reading can fail when coarse-graining and time reversal do not commute, Martínez et al. (2024) replied by restricting the original claim to local-in-time coarse-grainings and, where needed, second-order semi-Markov constructions, Blom et al. (2024) showed that coarse lumping can hide dissipative cycles and induce memory, and Baiesi et al. (2024) showed that sparse reverse transitions can force lower-bound strategies rather than direct estimation. Operational comparability is separate again: Poudel et al. (2024) showed motion-sensitive visibility-graph metrics with only selective moderate-to-high reliability in low-motion subsets, Metzen et al. (2024) showed that BOLD variability and complexity do not share one reliability profile, and Chen et al. (2025) showed that strong temporal coupling across simultaneous EEG-PET-MRI can coexist with non-identical spatial organization. Therefore, U10 now also separates reverse-transition support / finite-data handling, memory order / observed-state closure, stability / nuisance sensitivity, and physiology-bridge quality before any thermodynamic route is promoted.

For technical reading, split language route before calling it imitation

This section becomes too coarse if all language-facing outputs are treated as one `brain-to-text` bucket. Tang et al. (2023) constrain within-subject semantic reconstruction, Défossez et al. (2023) constrain fixed-segment retrieval, d'Ascoli et al. (2025) constrain known-onset word decoding, Ye et al. (2025) constrain prompt-conditioned generation, Willett et al. (2023), Littlejohn et al. (2025), and Wairagkar et al. (2025) constrain communication throughput, Card et al. (2024) and Singh et al. (2025) constrain decoder initialization, Pun et al. (2024) plus the fixed-decoder slice in Wairagkar et al. (2025) constrain fixed-decoder durability, and Karpowicz et al. (2025) plus Wilson et al. (2025) constrain adaptive rescue. These routes do not answer the same question, so this site reads them through the Neural Contribution Card before promoting any claim about mimic separation.