# Quantities registered with an estimator ladder nobody has built yet.
# Section 4.2: a quantity with no estimator is an open research target, and the docs
# build says so by name rather than treating it as a bug. This list may shrink and must
# not grow. Register an estimator and delete its line; docs/gen_catalogue.py --check
# enforces it, and --write-ledger rewrites it.
#
# Recorded from a base install, which is the leanest environment the docs build in. A
# richer environment imports more estimator modules and so sees fewer open quantities,
# which the ratchet allows.
#
# Five added at wave 6, and the reason is worth stating because this list may otherwise
# read as five new research problems. They are not. All five have a **shipped instrument
# that computes them**: the four `reward.*` are W3.9's contract layer, N5 to N8, built in
# wave 4, and `run.group_shape_factor` is what A1 reports beside its reading after the
# correction that stopped multiplying it in. What none of the five has is a registered
# `EstimatorEntry`, which is the ladder, so `best_estimator` cannot resolve them and the
# capability report cannot cost them. That is a real gap and it is the one this file is
# for; registering the five ladders is a wave-7 row. SPEC-ERRATA E56.
#
# This is the only legitimate way this list grows: a quantity registered without a ladder.
# The ratchet forbids growth that comes from an estimator being deleted or lost.
#
# One added at wave 7, and it is a second legitimate shape rather than a third:
# `policy.readout_recoverability` HAS a registered ladder, in `policy/quantities.py`, and that
# module imports `reward_lens.runtime`, which needs the `white-box` extra. So a base install
# never runs the registration and the quantity reads open here while being built and running
# everywhere torch is installed. That is the direction this file is recorded in on purpose: a
# richer environment sees strictly fewer open quantities, so a ledger written from the leanest
# one can only be too long, never too short. Register it outside the torch-gated import and
# this line goes away.
artifact.distillation_delta
baseline.best_score
baseline.margin
calibration.transfer_t32
credit.by_tool_call
credit.by_turn
credit.conservation_error
credit.implicit_prm
credit.localiser_quality
credit.measure
credit.successor_representation
env.attack_cheapness
env.attack_surface
env.flakiness
env.replay_fidelity
estimator.allfail_fraction
estimator.amplifier_safety
estimator.clip_fraction_effect
estimator.degenerate_fraction
estimator.noise_attribution
estimator.noise_share
estimator.spec
exploit.transfer_matrix
frontier.concomitant_bon
frontier.gold_vs_kl
frontier.prentice_checklist
frontier.tail_index
frontier.visibility_horizon
gate.bunching_elasticity
gate.deadzone_fraction
gate.mccrary_statistic
grader.abstention_rate
grader.allocation
grader.attenuation
grader.blackwell_order
grader.card
grader.circuit_overlap
grader.component_attribution
grader.component_patch_effect
grader.composition
grader.concept_dose_slope
grader.contested_axis
grader.counterfactual_score
grader.coverage_disparity
grader.criterion_coherence
grader.crystallization_depth
grader.curl_mass
grader.dark_fraction
grader.design_balance
grader.distortion_index
grader.effective_group_size
grader.eval_awareness
grader.feature_alignment
grader.flip_rate
grader.grr_percent
grader.harmonic_mass
grader.induced_variance
grader.knowledge_utilization
grader.legibility_frontier
grader.ndc
grader.null_reach
grader.objective_geometry
grader.path_effect
grader.prompt_snr
grader.receipt_reliance
grader.robustness_snr
grader.score_distribution
grader.score_sigma
grader.sep_distance
grader.silent_zero_rate
grader.skepticism
grader.style_share
grader.surface_bias_d
grader.tail_index
grader.term_conflict
grader.value_convergence_excess
grader.variance_components
grader.verification_score
instrument.effect
instrument.error_correlation
instrument.incremental_validity
instrument.interp_coverage
instrument.recovery_auc
instrument.shelf_life
intervention.acute_effect
intervention.att_staggered
intervention.chronic_effect
intervention.dissociation
intervention.ec50
intervention.emax
intervention.erasure_cost
intervention.hill_n
intervention.rescue_fraction
intervention.therapeutic_index
judge.commitment_position
judge.verdict_direction
labels.corrupt_success_rate
labels.error_rate
labels.false_success_rate
labels.fs_signal_locality
labels.position_prior
labels.score_ceiling
ladder.rung_disagreement
monitor.arl0
monitor.check_standard_drift
monitor.confidence_sequence
monitor.conjunction_alarm
monitor.degradation_curve
monitor.detection_delay
monitor.evalue
monitor.half_life
monitor.operating_point
monitor.ppv_curve
placebo.effect_ratio
policy.readout_recoverability
policy.train_infer_logprob_mismatch
policy.update_sparsity_vs_staleness
reference.u_characterisation
reference.u_homogeneity
reference.u_stability
reward.component_congruity
reward.equal_compensation_ratio
reward.optimal_weights
reward.sorting_cutoff
run.compaction_events
run.cot_faithfulness_under_kl
run.curriculum_response
run.group_shape_factor
run.importance_ess
run.mask_signature
run.max_staleness
run.variance_derivative
selection.G_conditioning
selection.autonomy
selection.covector
selection.differential_S
selection.dimensionality
selection.evolvability
selection.gradient_beta
selection.heritability_h2
selection.response_Gbeta
selection.spectrum
selection.stable_rank
study.mde
study.pooled_effect
study.power
study.prediction_interval_ratio
study.resolution_ratio
study.tau2
substrate.lod
substrate.loq
substrate.noise_floor
verifier.cfg
verifier.decision_coverage
verifier.false_positive_rate
verifier.fp_catalogue
verifier.metamorphic_violations
verifier.mutation_score
verifier.reliability_growth_beta
verifier.sobol_ST
verifier.surviving_mutants
verifier.trusted_inputs
verifier.unseen_exploit_mass
