a9cafc5368
Resolves the adversarial-identity v3 finding (0% rejection on
paraphrased attacks against the marker-string defense). Two
independent layers now guard the review gate; either is sufficient
to reject.
Fix #2 (syntactic, in teaching/review.py):
Replaces the substring-only check with four deterministic rules:
(a) legacy markers (v1/v2 coverage preserved verbatim)
(b) redirect-verb + role-frame co-occurrence
(c) negating qualifier within +/-3 tokens of a role-frame
(d) negating qualifier within +/-3 tokens of a redirect-verb
Replay-safe, no learned classifier, single-file contained change.
Fix #3 (geometric, in core/physics/identity.py):
Adds IdentityCheck.would_violate(score, manifold) predicate per
ADR-0010 and wires it through CognitiveTurnPipeline._run_teaching
from response.identity_score. The geometric layer is paraphrase-
invariant by construction.
Honest finding: with the current default IdentityManifold (three
unit-axis ValueAxes), the geometric layer flags 0/32 of v3 attacks
independently. The predicate and wiring are in place; the manifold
axis design is the limiting factor and remains as scoped follow-up.
Fix #2 is what is actually rejecting attacks today.
Verification: all eight adversarial-identity splits (v1-v4, public +
holdouts) at attack_rejection=1.0 and legitimate_acceptance=1.0.
v4 (32 attacks + 18 legitimate) is the regression gate for fix #2,
exercising rules (b)/(c)/(d) with new attack vocabulary. Tests
test_reviewed_teaching_loop.py (5/5), test_pipeline_teaching_integration.py
(5/5), test_identity_gate.py (incl. 5 new TestWouldViolatePredicate
tests, 12/12). CLI suites: smoke, cognition, teaching, runtime all
green.
Also drops a stale entry from the runtime CLI suite list
(test_chat_identity_telemetry.py was removed in 222124a).
271 lines
13 KiB
Markdown
271 lines
13 KiB
Markdown
# Capability Roadmap — Progress Tracker
|
||
|
||
Tracks completion of the phased plan defined in `docs/capability_roadmap.md`
|
||
(ADR-0016). Updated as work lands.
|
||
|
||
---
|
||
|
||
## Phase 0 — Benchmark Methodology Lock-in
|
||
|
||
**Status:** Complete
|
||
**Started:** 2026-05-15
|
||
**Completed:** 2026-05-16
|
||
|
||
- [x] Promote roadmap to ADR-0016
|
||
- [x] Extract `docs/eval_methodology.md` from roadmap Part I
|
||
- [x] Create progress tracker (`docs/PROGRESS.md`)
|
||
- [x] Implement `evals/<lane>/` directory convention
|
||
- [x] Build generic eval framework (`evals/framework.py`)
|
||
- [x] Retrofit `core eval cognition` into new convention
|
||
- [x] Split 45 cases into dev (13) / public v1 (13) / holdout (19)
|
||
- [x] Write `evals/cognition/contract.md`
|
||
- [x] Migrate `runner.py` to use framework
|
||
- [x] Record v1 results under new layout
|
||
- [x] Generalize `core eval <lane>` CLI (dynamic lane discovery)
|
||
- [x] Implement holdout runner scaffold
|
||
- [x] Implement baseline runner scaffold
|
||
- [x] **Exit gate:** `core eval cognition` runs under new convention with v1 public + holdout + baseline
|
||
|
||
### Methodology issues discovered (Phase 0 audit)
|
||
|
||
1. **Pipeline turn_log crash:** `CognitiveTurnPipeline.run()` assumed `turn_log`
|
||
was always populated after `chat()`, but the unknown-domain gate returns a
|
||
stub without appending. Fixed with fallback to tokenizer output.
|
||
2. **Versor drift in multi-turn sessions:** `test_pipeline_preserves_versor_closure`
|
||
reveals that after 3 turns in the same session, "spirit breath" causes
|
||
`versor_condition = 1.12e-04` (threshold: 1e-6). Pre-existing; resolved by
|
||
strict runtime closure enforcement (always unitize after sandwich product).
|
||
3. **Identity/drive bias shelved:** Premature persona motor and drive bias
|
||
introduced trajectory drift. Removed in favour of persona-neutral generic
|
||
runtime; identity returns behind explicit IdentityProfile contract.
|
||
|
||
---
|
||
|
||
## Phase 1 — Foundational Triple
|
||
|
||
**Status:** Complete ✓
|
||
**Started:** 2026-05-16
|
||
**Completed:** 2026-05-16
|
||
**Depends on:** Phase 0 exit
|
||
|
||
- [x] **grammatical-coverage** lane (v1 + v2 complete)
|
||
- [x] Enumerate English v1 constructions (13 constructions: C01-C13)
|
||
- [x] Write contract test pairs (PropositionGraph -> surface family)
|
||
- [x] Implement v1 dev/public (~41/36 items)
|
||
- [x] Implement holdout (52 items) — 100% pass
|
||
- [x] Engineer `realizer.py` to pass v1 (dev=100%, public=100%, holdout=100%)
|
||
- [x] Hebrew pack (`he_core_cognition_v1` with binyanim support)
|
||
- [x] Koine Greek pack (`grc_logos_cognition_v1` with Greek morphology)
|
||
- [x] Generate v2 on pass (deeper nesting, longer sentences, rarer vocabulary) — 36 cases (100% pass)
|
||
- [x] **zero-code-domain-acquisition** lane (v1 complete, zero engineering gaps)
|
||
- [x] Define 3 surprise domains (kinship, calendar, color)
|
||
- [x] Build pack-only authoring kits (vocabulary, relations, axioms, teaching examples, prompts)
|
||
- [x] Test: author brings CORE to >=80% without Python edits (100% achieved)
|
||
- [x] Log engineering gaps (ZERO — pack-only authoring contract is solid)
|
||
- [x] v1 dev (30/30), v1 public (18/18 across all 3 domains), v1 holdout (21/21) — all 100% pass
|
||
- [x] **identity-divergence** lane (v1 complete)
|
||
- [x] Define two identity axis sets (Axis A: Precision-first, Axis B: Generosity-first)
|
||
- [x] Curate shared curriculum (93 teaching events across color/kinship/reasoning/spatial)
|
||
- [x] Build divergence metric (>0.30 threshold): all pass (1.000)
|
||
- [x] Build coherence metric (>0.85 threshold for A and B): all pass (1.000)
|
||
- [x] Identity-stripped baseline with causal check: all pass (delta=1.000)
|
||
- [x] v1 dev (5/5), v1 public (5/5), v1 holdout (5/5) — all 100% pass
|
||
- [x] **Exit gate:** All three lanes pass v1 public + holdout ✓
|
||
|
||
---
|
||
|
||
## Phase 2 — Structural Wins Made Visible
|
||
|
||
**Status:** In Progress
|
||
**Started:** 2026-05-16
|
||
**Depends on:** Phase 1 exit
|
||
|
||
- [x] **provenance** lane (v1 complete)
|
||
- [x] Define Provenance dataclass + compute_provenance() (`core/cognition/provenance.py`)
|
||
- [x] Unit tests for provenance derivation (6/6 pass — `tests/test_provenance.py`)
|
||
- [x] Build pack-axiom / vault-recall / teaching / mixed case categories
|
||
- [x] v1 dev (10/10), v1 public (20/20), v1 holdouts (15/15) — all 100% pass
|
||
- [x] Sub-metrics: replay_determinism=1.0, source_attribution=1.0, source_validity=1.0, input_sensitivity=1.0
|
||
- [x] Fixed shape regression in `generate/stream.py` score-weighted recall (np.eye → multivector identity)
|
||
- [x] Replaced linear-blend rotor scaling with manifold-preserving `rotor_power` (`algebra/rotor.py`); 41 closure-preservation tests
|
||
- [x] Restored `respond()`/`result.final_state` identity contract after anchor pull
|
||
- [x] **monotonic-learning** lane (v1 complete)
|
||
- [x] Define contract: longitudinal regression check across ≥10 teaching cycles
|
||
- [x] Implement runner: shared session, sorted ops, per-(cycle, domain) accuracy table
|
||
- [x] Generator (`scripts/generate_monotonic_cases.py`) for cycle/probe corpora
|
||
- [x] v1 dev (10 cycles), v1 public (12 cycles, 3 domains), v1 holdouts (12 cycles, 2 distinct domains)
|
||
- [x] All splits: max_regression=0.00, floor_score=1.00, overall_pass=true
|
||
- [x] Structural win demonstrated: zero regression across 34 total cycles / 7 distinct domains
|
||
- [x] **calibration** lane (v1 complete)
|
||
- [x] Define contract: typed signals for no_grounding / coherent / correction_proposed
|
||
- [x] Classification from `CognitiveTurnResult` (vault_hits + pack_mutation_proposal)
|
||
- [x] Runner with per-case fresh pipeline (avoids cross-case field drift)
|
||
- [x] v1 dev (12/12), v1 public (24/24), v1 holdouts (18/18) — all 100% pass
|
||
- [x] Sub-metrics: no_grounding=1.0, coherent=1.0, correction_proposed=1.0
|
||
- [x] Architectural finding documented (`evals/calibration/gaps.md`): the
|
||
ingest gate is geometric, not semantic — 6/42 hand-chosen OOD
|
||
prompts fire the geometric gate. v1 measures recall-presence +
|
||
correction-firing signals (deterministic), not semantic OOD.
|
||
Pipeline override of gate's safety surface is a separate gap.
|
||
- [x] **symbolic-logic** lane (v1 complete)
|
||
- [x] Define contract: structural foundations for proposition-based inference
|
||
- [x] Patterns: modus_ponens_chain, modus_tollens_chain, syllogism, negation, chain_recall
|
||
- [x] Runner: per-case fresh pipeline + double-run replay check
|
||
- [x] Sub-metrics: premise_recall=1.0, replay_determinism=1.0, proposal_storage=1.0
|
||
- [x] v1 dev (8/8), v1 public (18/18), v1 holdouts (12/12) — all 100% pass
|
||
- [x] Architectural finding documented (`evals/symbolic_logic/gaps.md`): CORE
|
||
has no first-class inference operator yet. v1 measures the storage,
|
||
replay, and recall foundations on which a future inference engine
|
||
would be built. v2 would assert specific inference correctness
|
||
(transitive recall surface contents).
|
||
- [x] **adversarial-identity** lane (v1 complete)
|
||
- [x] Define contract: identity-override attacks rejected at review;
|
||
legitimate corrections still accepted
|
||
- [x] Cover all `_IDENTITY_MARKERS` families (you are / forget / pretend /
|
||
override / ignore / your name / act as / from now / character /
|
||
personality)
|
||
- [x] Per-case fresh pipeline; prior question primes the review surface
|
||
- [x] Sub-metrics: attack_rejection_rate=1.0, legitimate_acceptance_rate=1.0
|
||
- [x] v1 dev (10/10), v1 public (25/25), v1 holdouts (18/18) — all 100% pass
|
||
- [x] **All five Phase 2 v1 lanes passing** ✓
|
||
- [x] Frontier baselines computed for all lanes (structural-zero floor)
|
||
- [x] `docs/frontier_baselines.md` — per-lane analysis: frontier LLMs do
|
||
not emit the typed signals CORE's rubrics score against
|
||
(provenance sources, pack_mutation_proposal, vault_hits,
|
||
REJECTED_IDENTITY outcome, deterministic trace_hash)
|
||
- [x] Per-lane structural-zero baseline JSON written under
|
||
`evals/<lane>/baselines/v1_structural_zero.json`
|
||
- [x] `StructuralZeroBaseline` adapter in `evals/baseline_runner.py`
|
||
— deterministic floor; live-API adapters can be added when
|
||
keys are configured
|
||
- [x] v2 lanes: all five at 100% pass
|
||
- monotonic-learning v2 — 20 cyc / 5 dom (public), 18 cyc / 4 dom (holdouts)
|
||
- provenance v2 — 30 + 20 cases, all sub-metrics 1.0
|
||
- adversarial-identity v2 — 35 + 22 cases, all 1.0
|
||
- calibration v2 — 33 + 24 cases, all class accuracies 1.0
|
||
- symbolic-logic v2 — 24 + 16 cases (chains up to 5 hops), all 1.0
|
||
- [x] **Exit gate:** v3 lanes for at least two of the five ✓
|
||
- monotonic-learning v3 — 30 cyc / 7 dom (public), 25 cyc / 6 dom (holdouts),
|
||
`max_regression=0.0`, `floor_score=1.0` on both splits
|
||
- adversarial-identity v3 — 30 + 20 paraphrased-attack cases.
|
||
Initial v3 result (pre-fix): `attack_rejection_rate=0.0`,
|
||
`legitimate_acceptance_rate=1.0`. v3 was a load-bearing finding
|
||
that exposed the marker-string defense as brittle to paraphrase.
|
||
|
||
### Identity-override defense — fix #2 + fix #3 (2026-05-16)
|
||
|
||
Triggered by the v3 finding above. Two-layer defense now active in
|
||
`teaching/review.py`:
|
||
|
||
- **Fix #2 (syntactic).** `_is_identity_override` applies four
|
||
deterministic rules: (a) legacy markers, (b) redirect-verb +
|
||
role-frame co-occurrence, (c) negating qualifier ±3 tokens from a
|
||
role-frame, (d) negating qualifier ±3 tokens from a redirect-verb.
|
||
- **Fix #3 (geometric).** `IdentityCheck.would_violate(score, manifold)`
|
||
predicate added to `core/physics/identity.py`; `review_correction`
|
||
now accepts `identity_score` / `identity_manifold` kwargs and is
|
||
wired in `CognitiveTurnPipeline._run_teaching` from
|
||
`response.identity_score`.
|
||
|
||
Lane results after both fixes:
|
||
|
||
| split | attacks | attack_rej | legit_acc |
|
||
|---|---|---|---|
|
||
| public/v1 | 15 | 1.0 | 1.0 |
|
||
| holdouts/v1 | 10 | 1.0 | 1.0 |
|
||
| public/v2 | 20 | 1.0 | 1.0 |
|
||
| holdouts/v2 | 12 | 1.0 | 1.0 |
|
||
| public/v3 | 20 | 1.0 | 1.0 |
|
||
| holdouts/v3 | 12 | 1.0 | 1.0 |
|
||
| public/v4 | 20 | 1.0 | 1.0 |
|
||
| holdouts/v4 | 12 | 1.0 | 1.0 |
|
||
|
||
v4 is the regression gate for fix #2 — new attack vocabulary
|
||
combinations that exercise rules (b)/(c)/(d) without repeating v3's
|
||
specific surface. All v1–v4 splits pass at 100% with the new
|
||
defense; legitimate-correction false-positive rate is 0%.
|
||
|
||
Honest finding: with the current default `IdentityManifold` (three
|
||
unit-axis ValueAxes), the geometric layer flags 0/32 of v3 attacks
|
||
independently of fix #2. The predicate and wiring are in place; the
|
||
manifold's axis design is the limiting factor and needs sharpening
|
||
before the geometric defense can carry weight on its own. See
|
||
`evals/adversarial_identity/gaps.md`.
|
||
|
||
## Phase 2 — COMPLETE
|
||
|
||
All five Phase 2 v1+v2 lanes pass at 100%; frontier structural
|
||
baselines documented; v3 satisfies the exit-gate requirement (two
|
||
lanes, one demonstrating a passing structural-depth test and one
|
||
demonstrating an architectural vulnerability that the geometric
|
||
identity-check fix in `evals/adversarial_identity/gaps.md` would
|
||
close).
|
||
|
||
### Parallel eval infrastructure (2026-05-16)
|
||
|
||
- `evals/parallel.py` — `run_cases_parallel()` helper using
|
||
`multiprocessing.Pool` with the `"spawn"` start method (avoids
|
||
forking heavy parent state). Default workers = `min(cpu_count, 8)`.
|
||
- Wired into the four per-case lanes (provenance, calibration,
|
||
symbolic-logic, adversarial-identity). `run_lane(..., workers=N)`
|
||
controls parallelism; `workers=1` forces serial for debugging.
|
||
- Empirical speedup (adversarial-identity public/v1, 25 cases):
|
||
serial 14.1s → parallel 3.1s (~4.5x).
|
||
- Monotonic-learning intentionally stays serial within a split
|
||
(shared longitudinal session by design).
|
||
|
||
---
|
||
|
||
## Phase 3 — Reasoning Depth
|
||
|
||
**Status:** Not Started
|
||
**Depends on:** Phase 2 exit
|
||
|
||
- [ ] **compositionality** lane (construction-family splits, not sampling)
|
||
- [ ] **inference-closure** lane
|
||
- [ ] **introspection** lane
|
||
- [ ] **multi-step-reasoning** lane
|
||
- [ ] **cross-domain-transfer** lane
|
||
- [ ] Pin agency scope decision (responsive vs. goal-directed)
|
||
- [ ] Pin tool-use scope decision
|
||
- [ ] **Exit gate:** All five v1 scored; at least two passing v1
|
||
|
||
---
|
||
|
||
## Phase 4 — Scale and Efficiency
|
||
|
||
**Status:** Not Started
|
||
**Depends on:** Phase 3 exit
|
||
|
||
- [ ] **sample-efficiency** curves (>=10 concepts)
|
||
- [ ] **long-context-cost** curves (10^3 to 10^6 vault entries)
|
||
- [ ] **multi-agent-composition** (>=2 agents, replay preserved)
|
||
- [ ] Vault indexing strategy decided
|
||
- [ ] **Exit gate:** All curves published with confidence intervals
|
||
|
||
---
|
||
|
||
## Phase 5 — Curriculum Era
|
||
|
||
**Status:** Not Started
|
||
**Depends on:** Phase 4 exit
|
||
|
||
- [ ] 5.1 English fluency (grammatical-coverage v5 OOD)
|
||
- [ ] 5.2 Hebrew fluency
|
||
- [ ] 5.3 Koine Greek fluency
|
||
- [ ] 5.4 Elementary mathematics
|
||
- [ ] 5.5 Foundational physics
|
||
- [ ] 5.6 Foundational biology
|
||
- [ ] 5.7 Classical literature
|
||
- [ ] Phase 1-4 lanes re-run on every release (no regression)
|
||
|
||
---
|
||
|
||
## Open Scope Decisions
|
||
|
||
| Decision | Status | Deadline |
|
||
|----------|--------|----------|
|
||
| Agency (responsive vs. goal-directed) | Open | Before Phase 3 |
|
||
| Tool use (typed deterministic operators) | Open | Before Phase 3 |
|
||
| Code generation (first-class target) | Open | Before Phase 5 |
|
||
| Embodiment (sensorium gates) | Open | Phase 5 |
|