core-labs/core
1
0
Fork 0
Code Issues Pull requests 1 Projects Releases Packages Wiki Activity Actions
core/tests
History
Shay b6649848a6 feat(adr-0024): Phase 5 — stratified mechanism-isolation across 5 failure-mode families 
Authors a 20-case corpus stratified across five geometric failure-mode
families and a separate 10-case benign corpus for the
EXHAUSTION_CEILING lane:

  A. near_forbidden_correct_endpoint  (6 cases, gaps 0.002 to 0.55)
  B. near_equal_admissible             (5 cases, diffs ≤ 0.01)
  C. no_admissible_path                (3 cases, honest refusal)
  D. multi_step_admissibility          (3 chained cases)
  E. heterogeneous_relation            (3 chained cases, blade-switching)

phase5_runner runs each case under BOTH threshold and ADR-0026 margin
modes and reports per-family pass_rate, refusal_rate, and (for Family
A) rejection_traced_rate + boundary_overridden_rate.

Headline:
  pass_rate_threshold = 1.00 (20/20)
  pass_rate_margin    = 1.00 (20/20)
  mechanism_isolated  = true (both modes, all five families)
  replay determinism  = byte-identical across 3 reruns

Family C refuses with RefusalReason.INNER_LOOP_EXHAUSTION in both
modes (load-bearing evidence for ADR-0024 Phase 2 typed refusals).
Family B refuses under margin mode (validates ADR-0026 δ=0.4 gate).

Benign inner-loop corpus for EXHAUSTION_CEILING ≤ 0.05 gate:
  boundary_only:    exhaustion 0.00, pass 1.00
  null_control:     exhaustion 0.00, pass 1.00
  inner_loop_t0:    exhaustion 0.00, pass 1.00
  inner_loop_tpos:  exhaustion 0.00, pass 1.00 (threshold 0.25)

Geometric finding documented while authoring the benign corpus:
23 of 85 pack tokens have negative self-cga_inner under Cl(4,1).
Tokens with self-score ≤ 0 cannot serve as single-token expected
endpoints in threshold mode — the algebra's Lorentzian signature
forbids this geometrically.  Phase 5 benign corpus draws expected
endpoints from the 62-token positive-self-score subset.  This is
consistent with Phase 4 characterization: no static threshold
delivers separation_quality ≥ 0.8 — the margin lane survives
because margin compares differences, not absolute scores.

Files:
  evals/forward_semantic_control/public/v2_phase5/cases.jsonl
  evals/forward_semantic_control/public/inner_loop_benign/cases.jsonl
  evals/forward_semantic_control/phase5_runner.py
  evals/forward_semantic_control/phase5_mine.py
  evals/forward_semantic_control/results/phase5_report.json
  evals/forward_semantic_control/results/phase5_benign_inner_loop_report.json
  tests/test_phase5_corpus.py        (20 passing)
  docs/evals/phase5_stratified_findings.md

Tests: 1068 passed, 2 skipped (+20 from Phase 4 baseline).
2026-05-17 15:51:59 -07:00
..
formation feat(epistemic): truth-seeking schema audit — 3 leaks closed, 4 new lanes, 3 new invariants 2026-05-17 07:27:41 -07:00
__init__.py init: tests, pyproject.toml, AGENTS.md, CLAUDE.md, README.md 2026-05-12 19:15:28 -07:00
test_achat.py test(chat): achat() and arespond() smoke tests — surface, fields, OOV grace, structural determinism 2026-05-14 14:31:49 -07:00
test_admissibility_trace.py feat(adr-0023): Forward Semantic Control proof evidence — Accepted 2026-05-17 12:55:19 -07:00
test_alignment_graph.py Fix test suite errors across core physics and generation 2026-05-14 13:02:32 -07:00
test_architectural_invariants.py feat(epistemic): Leak C read-side audit — INV-24 callsite registry, Leak C fully closed 2026-05-17 09:48:39 -07:00
test_articulation.py fix(tests): align suite with persona-neutral strict-closure contract 2026-05-16 05:37:12 -07:00
test_articulation_realizer_v2.py Add articulation realizer v2 2026-05-14 20:14:50 -07:00
test_benchmarks_profiler.py feat(benchmarks): discourse_paragraph lane + pipeline profiler + word-selection tracer 2026-05-16 21:53:46 -07:00
test_cga_inner_rust_parity.py feat(adr-0020): parity gates for cga_inner, geometric_product, versor_condition, versor_apply 2026-05-16 20:37:58 -07:00
test_chat_runtime.py Bridge chat OOV grounding 2026-05-13 20:47:28 -07:00
test_cli.py fix(tests): align suite with persona-neutral strict-closure contract 2026-05-16 05:37:12 -07:00
test_cli_pack_validate_security.py Harden pack validator trust boundary 2026-05-15 11:50:15 -07:00
test_cli_test_suites.py Speed up validation lanes and pack loading 2026-05-15 08:26:11 -07:00
test_cognitive_eval_harness.py Add cognitive eval harness and calibration replay (#30) 2026-05-15 07:41:36 -07:00
test_cognitive_turn_pipeline.py Wire intent graph into cognitive pipeline 2026-05-14 20:05:00 -07:00
test_compiler.py feat(core_ingest): IngestPipeline, manifold integration, full test suite 2026-05-13 12:07:12 -07:00
test_compose_relations.py feat(compositionality): compose_relations operator lifts lane 68.8% → 100% 2026-05-16 22:44:06 -07:00
test_core_ingest.py feat: scaffold core_ingest/ governance layer (ADR-0012) 2026-05-13 11:35:00 -07:00
test_core_semantic_seed_pack.py feat(packs): en_core_cognition_v1 v1.2.0 - rhetoric/metaphor/narrative 2026-05-16 15:15:14 -07:00
test_determinism_proofs.py Restore FieldState slots in determinism proof 2026-05-13 14:26:24 -07:00
test_deterministic_hash.py feat: manifold field topology, graph diffusion operator, vertical pulse 2026-05-15 16:02:48 -07:00
test_dialogue.py fix(tests): align suite with persona-neutral strict-closure contract 2026-05-16 05:37:12 -07:00
test_dialogue_fluency_regression.py fix: harden session field invariants and eliminate hot-path inefficiencies 2026-05-15 21:37:49 -07:00
test_energy.py Tighten hot-path backend consistency 2026-05-15 08:14:38 -07:00
test_engine_loop_proof.py Tighten session node tracking and generation selection 2026-05-13 14:35:31 -07:00
test_epistemic_invariants.py feat(adr-0021): epistemic_status surface wired across teaching + trace 2026-05-16 20:20:35 -07:00
test_explain.py feat(phase3): core/cognition/explain.py — close Gap 3 introspection 2026-05-16 15:09:48 -07:00
test_forward_semantic_control.py feat(adr-0022): Forward Semantic Control — Accepted 2026-05-17 12:10:20 -07:00
test_generate_stream.py fix(tests): align suite with persona-neutral strict-closure contract 2026-05-16 05:37:12 -07:00
test_geometric_product_rust_parity.py feat(adr-0020): parity gates for cga_inner, geometric_product, versor_condition, versor_apply 2026-05-16 20:37:58 -07:00
test_graph_diffusion.py feat: manifold field topology, graph diffusion operator, vertical pulse 2026-05-15 16:02:48 -07:00
test_holonomy.py Align holonomy tests with indefinite metric 2026-05-13 12:59:32 -07:00
test_holonomy_resonance.py Implement trilingual field coherence 2026-05-13 19:53:37 -07:00
test_identity_gate.py fix(identity): close v3 paraphrase gap with two-layer override defense 2026-05-16 14:05:55 -07:00
test_inference_operators.py feat(phase3): multi_relation_walk closes Phase 3 v1 to 10/10 splits 2026-05-16 15:24:44 -07:00
test_inner_loop_admissibility.py feat(adr-0024): Phases 2-5 — corpus eval, v2 adversarial, threshold characterization, ADR-0025 design note 2026-05-17 14:07:50 -07:00
test_inner_loop_phase2.py feat(adr-0024): Phases 2-5 — corpus eval, v2 adversarial, threshold characterization, ADR-0025 design note 2026-05-17 14:07:50 -07:00
test_inner_loop_phase3.py feat(adr-0024): Phases 2-5 — corpus eval, v2 adversarial, threshold characterization, ADR-0025 design note 2026-05-17 14:07:50 -07:00
test_inner_loop_phase4.py feat(adr-0024): Phase 1 addendum — retire v1/dev fixture rot 2026-05-17 14:43:34 -07:00
test_intent_proposition_graph.py feat: add intent-proposition graph comprehension layer 2026-05-14 19:52:57 -07:00
test_intent_ratifier.py feat(adr-0022): Forward Semantic Control — Accepted 2026-05-17 12:10:20 -07:00
test_language_pack_cache.py feat(adr-0024): Phase 1 — wire inner-loop admissibility + determinism proof 2026-05-17 13:38:55 -07:00
test_language_pack_runtime.py Implement trilingual field coherence 2026-05-13 19:53:37 -07:00
test_manifold.py feat(core_ingest): IngestPipeline, manifold integration, full test suite 2026-05-13 12:07:12 -07:00
test_manifold_state.py feat: manifold field topology, graph diffusion operator, vertical pulse 2026-05-15 16:02:48 -07:00
test_margin_admissibility.py feat(adr-0026): Phase 3 — ranked admissibility with margin 2026-05-17 15:03:03 -07:00
test_morphology_irregular.py fix(gaps): close G1+G2+G3 + identity vocab + pipeline safety-stub honour 2026-05-16 21:21:06 -07:00
test_morphology_registry.py Implement trilingual field coherence 2026-05-13 19:53:37 -07:00
test_motor.py Fix full test suite after cognitive runtime 2026-05-13 13:52:11 -07:00
test_null_cone.py init: tests, pyproject.toml, AGENTS.md, CLAUDE.md, README.md 2026-05-12 19:15:28 -07:00
test_oov_grounding_cache.py Cache OOV morphology grounding structures 2026-05-15 11:53:46 -07:00
test_operator_calibration_replay.py Add cognitive eval harness and calibration replay (#30) 2026-05-15 07:41:36 -07:00
test_phase5_corpus.py feat(adr-0024): Phase 5 — stratified mechanism-isolation across 5 failure-mode families 2026-05-17 15:51:59 -07:00
test_pipeline.py feat(core_ingest): IngestPipeline, manifold integration, full test suite 2026-05-13 12:07:12 -07:00
test_pipeline_teaching_integration.py feat: wire teaching loop into cognitive pipeline 2026-05-14 20:46:50 -07:00
test_probe_repl.py Implement trilingual field coherence 2026-05-13 19:53:37 -07:00
test_proof_properties.py feat: Full Proof — surface realizer join, Rust diffusion parity, benchmark harness 2026-05-15 17:39:14 -07:00
test_proposition.py Add proposition generation 2026-05-13 20:08:49 -07:00
test_provenance.py feat(evals): provenance lane v1 — replay determinism + source back-pointers 2026-05-16 11:45:00 -07:00
test_pulse_integration.py feat: Full Proof — surface realizer join, Rust diffusion parity, benchmark harness 2026-05-15 17:39:14 -07:00
test_realizer_quantifier_agreement.py fix(gaps): close G1+G2+G3 + identity vocab + pipeline safety-stub honour 2026-05-16 21:21:06 -07:00
test_refusal_contract.py feat(adr-0024): Phase 2 — honest refusal with typed evidence 2026-05-17 14:49:08 -07:00
test_replay_vs_llm_benchmark.py feat(compositionality): compose_relations operator lifts lane 68.8% → 100% 2026-05-16 22:44:06 -07:00
test_reviewed_teaching_loop.py feat: add reviewed teaching loop for controlled correction learning 2026-05-14 20:32:28 -07:00
test_rotor_admissibility.py feat(adr-0025): Phase 4 — rotor / frame admissibility at the seam 2026-05-17 15:16:32 -07:00
test_rotor_power.py fix(drift): proper rotor-manifold scaling; restore respond contract 2026-05-16 11:44:45 -07:00
test_runtime_config.py fix(tests): align suite with persona-neutral strict-closure contract 2026-05-16 05:37:12 -07:00
test_rust_backend.py feat(algebra): null-preserving versor_apply path + un-skip 2 invariant tests 2026-05-16 21:40:37 -07:00
test_salience.py fix(tests): align suite with persona-neutral strict-closure contract 2026-05-16 05:37:12 -07:00
test_segmenter.py feat(core_ingest): IngestPipeline, manifold integration, full test suite 2026-05-13 12:07:12 -07:00
test_semantic_realizer_integration.py feat(adr-0023): Forward Semantic Control proof evidence — Accepted 2026-05-17 12:55:19 -07:00
test_sensorium_mount.py feat: scaffold sensorium/ (ADR-0013) + tests/test_architectural_invariants.py + tests/test_sensorium_mount.py 2026-05-13 11:45:16 -07:00
test_session_coherence.py fix(session): CGA-only hemisphere consistency, remove forbidden Euclidean metric 2026-05-15 21:51:16 -07:00
test_surface.py test(surface): full unit test suite for SentenceAssembler — roles, guard, elaboration, languages, determinism 2026-05-14 14:29:46 -07:00
test_transition_rotor.py fix(algebra): enforce transition rotor invariants 2026-05-15 21:13:14 -07:00
test_unknown_token_ingest.py Ground unknown tokens in ingest 2026-05-13 20:33:20 -07:00
test_vault_config.py fix(tests): align suite with persona-neutral strict-closure contract 2026-05-16 05:37:12 -07:00
test_vault_recall.py Activate and verify Rust backend 2026-05-13 22:23:48 -07:00
test_vault_recall_rust_parity.py feat(rust-parity): vault_recall first surface — bit-identity gate passes 2026-05-16 17:14:22 -07:00
test_vault_recall_vectorised.py feat(phase4): long-context-cost lane + ADR-0019 Stage 1 vault recall vectorisation 2026-05-16 16:39:30 -07:00
test_vault_store.py feat: vault recall index, Rust versor parity, cognitive pack expansion 2026-05-15 15:34:39 -07:00
test_versor_apply_rust_parity.py perf(rust): versor_apply f64 parity port — 29x over Python, bit-identical 2026-05-16 20:43:01 -07:00
test_versor_closure.py feat(algebra): null-preserving versor_apply path + un-skip 2 invariant tests 2026-05-16 21:40:37 -07:00
test_versor_condition_rust_parity.py feat(adr-0020): parity gates for cga_inner, geometric_product, versor_condition, versor_apply 2026-05-16 20:37:58 -07:00
test_vocab_manifold_invariants.py Fail closed on invalid versor construction 2026-05-14 10:55:11 -07:00