task_id	name	generalization_axis	state	input	output	outcome_required	permanent_baselines	primary_metrics	current_result	decision_blocker	route
G0	Same-context target holdout	unseen target inside the processed HepG2 assay	EXECUTED_LIMITED	Processed HepG2 CRISPRi response deltas	Held-out 2,000-feature mean response vector	Released same-assay target aggregates	zero change · train mean · ridge	RMSE · raw cosine · residual cosine · strict 16-outcome gate	Small average-error gain; strict gate 0/16 and direction unreliable	No external cell-level or disease-context truth	/resource/benchmarks/core/
G1	External unseen perturbation	new perturbation in an external aggregate object	EXECUTED_PARTIAL_UNSUPPORTED	HepG2 response substrate plus 55-target external aggregate diagnostic	Directional-transfer assessment	Frozen aggregate target effects	zero change · train mean · ridge	formula-aligned aggregate directional diagnostics	No directional-transfer support	No harmonized cell-population outcome or matched feature space	/resource/benchmarks/external-transfer/
G2	Unseen gene family or pathway	held-out related targets with leakage-safe exposure audit	REGISTERED_NOT_EXECUTED	Future exposure-audited perturbation training set	Held-out family response distribution	Enough related targets to freeze disjoint train and test families	zero change · train mean · ridge · nearest neighbour	DES · PDS · MAE · calibration · abstention	No result	Current candidate set is too small for a powered family holdout	/resource/benchmarks/generalization/
G3	Unseen donor	entire donor withheld	LOCKED_NO_TRUTH	Future donor-resolved perturbation populations	Donor-conditioned response and responder distribution	Multiple biological donors with repeated perturbations	population mean · donor-agnostic ridge · nearest donor	calibration · donor replication · distribution distance · abstention	No result	No qualifying donor-resolved DMD perturbation matrix	/resource/benchmarks/generalization/
G4	Unseen cell state	train in one myogenic state and test another	LOCKED_NO_TRUTH	Future perturbations sampled across myoblast-to-myotube states	State-conditioned population transition	Matched perturbation identities across physical time and state	unperturbed trajectory · state mean · additive shift	state occupancy · DES · PDS · MAE · transport distance	No result	Current myogenic trajectory is real but unperturbed	/resource/trajectory/
G5	Future batch or external laboratory	prospective time and site shift	LOCKED_NO_TRUTH	Future frozen prediction plus later independent batch	Prospective response, uncertainty and abstention	Outcome collected after Prediction Card freeze	historical mean · ridge · expert selection	calibration · AUPRC · hit enrichment · abstention quality	No result	No registered prediction or future outcome	/resource/prediction_registry
G6	Independent DMD disease line	new disease line or genotype	LOCKED_NO_TRUTH	Future matched DMD/control perturbation conditions	Molecular, functional, toxicity and replication outcomes	Independent DMD candidate perturbations across donors and reagents	zero · train mean · ridge · expert choice · random selection	functional hit rate · AUPRC · calibration · toxicity miss rate · replication	No result	0 independent DMD candidate-perturbation outcomes	/resource/minimum-perturbome/
G7	New modality and combination	KO, CRISPRa, drug or double perturbation	LOCKED_NO_TRUTH	Future matched singleton and combination perturbations	Main effects, interaction, synergy class and uncertainty	Observed same-screen singles and doubles or matched genetic–chemical pairs	additive null · pathway mean · nearest neighbour	interaction error · synergy classification · calibration · toxicity	Only an additive no-interaction preview is available	No observed candidate doubles or matched DMD drug screens	/resource/perturbation-lab