watch_id	model_name	release_date	evidence_state	model_family	reported_scope	openness	local_state	local_result	priority	nmd_evidence	next_gate	sources
WATCH-TXPERT-2026	TxPert	2026-05-01	PEER_REVIEWED_PRIMARY_AND_OFFICIAL_CODE	knowledge-graph perturbation model	Unseen single and double perturbations plus cross-cell-line transcriptomic response.	Paper and official reproduction repository available; exact asset licences must be checked per artifact.	EXECUTED_LOCAL_GATE_FAIL	The released NMD-VCell seed did not beat train mean on its frozen HepG2 gate; confirmatory seeds remain locked.	P0_ALREADY_EXECUTED_NO_ADVANCEMENT	NOT_ESTABLISHED	No repeat by reputation; require a new preregistration, compatible truth or untouched test partition.	https://www.nature.com/articles/s41587-026-03113-4 | https://github.com/valence-labs/TxPert
WATCH-STATE-2025	STATE	2025-06-01	OFFICIAL_SOURCE_AND_CODE_VERIFIED	set-based state transition model	Population-level perturbation response across experimental contexts.	Official code and model ecosystem available; licence checked per checkpoint.	EXECUTED_8_LEGACY_CONFIGS_ALL_GATE_FAIL	Eight CPU/GPU and held-out/seen-perturbation receipts were migrated; none beat both zero and train mean on every split.	P1_EXECUTED_NO_ADVANCEMENT	NOT_ESTABLISHED	Do not rerun by reputation; require a corrected preregistration, untouched truth or materially different task-compatible checkpoint.	https://github.com/ArcInstitute/state | https://arcinstitute.org/virtual-cell-initiative
WATCH-STACK-2026	Stack	2026-01-09	OFFICIAL_SOURCE_PREPRINT_AND_CODE_VERIFIED	in-context single-cell foundation model	In-context representation and condition transfer; official source reports pretraining on 149 million cells.	Official source links preprint, code and Perturb Sapiens artifacts.	LOCAL_NVME_CORE_SMOKE_PASS_NO_COMPATIBLE_CHECKPOINT	After one 180-second dependency-path timeout, a bounded local-NVMe retry passed the official StateICLModelBase forward path on a V100 with finite outputs. No pretrained checkpoint or task benchmark was run.	P1_BLOCKED_TASK_COMPATIBLE_CHECKPOINT	NOT_ESTABLISHED	Identify a task-compatible pretrained genetic checkpoint and preregister an untouched benchmark before any model-performance claim.	https://github.com/ArcInstitute/stack | https://arcinstitute.org/news/foundation-model-stack
WATCH-TAHOE-X1-2025	Tahoe-x1	2025-10-24	OFFICIAL_MODEL_CARD_AND_WEIGHTS_VERIFIED	single-cell foundation model	70M, approximately 1B and 3B checkpoints linked to Tahoe-100M.	Official Hugging Face model card lists Apache-2.0 and public weights.	SOURCE_PINNED_INPUT_CONTRACT_FAIL	The frozen 145,473-cell H5AD is log-normalized, has no raw layer and therefore fails Tahoe-x1's raw-count input contract; no model run was claimed.	P1_BLOCKED_RAW_COUNT_INPUT	NOT_ESTABLISHED_CANCER_TO_MUSCLE_SHIFT	Acquire count-preserving input with auditable gene identifiers; never exponentiate normalized values and relabel them raw counts.	https://huggingface.co/tahoebio/Tahoe-x1
WATCH-SLIM-2026	SLIM	2026-08-07	RESEARCH_REPORT_LEAD_PRIMARY_SOURCE_PENDING	small linear model with biological priors	The supplied research report describes a STRING-informed linear perturbation comparator.	Not accepted into the source-verified registry until a primary paper and code identity are pinned.	NOT_ELIGIBLE_FOR_EXECUTION	No source-verified artifact and no local run.	P1_HOLD_SOURCE_VERIFICATION	NOT_ESTABLISHED	Verify primary source, code, licence and exact task before creating an adapter.	
