Learned tracking and agent models examine rapid mouse route choice
Source check: Source checked. Not independent scientific confirmation.
Finding: Neural tracking supported evidence that mice favor shorter routes and smaller shelter angles before escape experience. Model-based learning agents reproduced rapid route choice from recorded exploration, whereas the tested model-free agent failed under that limited experience.
Technology: DeepLabCut neural pose estimation; model-based reinforcement learning
Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.
- AI contribution
- A learned pose network extracts movement; reinforcement-learning agents test algorithmic explanations of recorded behavior.
- Test and result
- Custom-trained pose estimation measured trajectories. Q-learning, DYNA-Q and Influence Zones agents were compared using42 retained exploration sessions.
- Limitation / still unknown
- Agent agreement does not identify the biological learning algorithm. Recorded sessions and repeated trials are not independent neural-mechanism tests; alternative prior-knowledge explanations remain. No causal brain mechanism is claimed.
Evidence sources and annotation context
Evidence context (descriptions follow the cited sources; test type is a catalog classification):
Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC9616796/
Source checked · editorial details
The source page was checked against this entry. Independent confirmation is not assessed by this label.
Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.
Review date: 21 Nov 2026. Review scope: Safe tracking/RL computational Methods and high-level Results/Discussion; primary date/status. No operational animal procedure needed for represented scope..
No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.
Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.