Shay 819c8b81ac feat(phase3): compositionality, multi-step-reasoning, introspection, cross-domain-transfer v1

Spreads the four remaining Phase 3 lanes to map the full reasoning-
depth surface alongside inference-closure (already landed at e509e0d).
Each lane is a v1 honest probe per the roadmap; engineering work
follows once the full surface is visible.

Results across all five Phase 3 lanes:

  lane                      split        primary signal  foundation
  inference-closure         public/v1    0.0             1.0 / 1.0
  inference-closure         holdouts/v1  0.0             1.0 / 1.0
  compositionality          public/v1   0.0625 (1/16)   1.0 / 1.0
  compositionality          holdouts/v1  0.0             1.0 / 1.0
  multi-step-reasoning      public/v1    0.0             1.0 / 1.0
  multi-step-reasoning      holdouts/v1  0.0             1.0 / 1.0
  introspection             public/v1    0.0 (no api)    n/a
  introspection             holdouts/v1  0.0             n/a
  cross-domain-transfer     public/v1    0.0             1.0 / 1.0
  cross-domain-transfer     holdouts/v1  0.0             1.0 / 1.0

Foundation guarantees (storage + replay) intact across every lane
that has them. The reasoning-depth signal is uniformly zero. The
five lanes triangulate four architectural gaps:

  Gap 1. generate/graph_planner.py has no transitive composition.
  Gap 2. field/propagate.py has no derivable-but-not-asserted recall.
  Gap 3. core/cognition/explain.py module does not exist.
  Gap 4. no structural-pattern recogniser (cross-subdomain transfer).

Gaps 1, 2, 4 cluster on the same code surface and may close together
as a single bounded PR. Gap 3 is independent module-creation work.

Lane scaffolding mirrors inference-closure (contract.md, runner.py,
dev + public/v1 + holdouts/v1 cases.jsonl, baselines/v1_structural_zero.json,
gaps.md). All runners are parallel-safe and use the standard
run_lane(cases, *, config, workers) interface.

Per-lane gaps.md records the engineering shape for v2 plus future
directions worth not forgetting:
  - compositionality/gaps.md: metaphor is compositionality with
    selective property transfer; building it is correctly downstream
    of closing this lane.
  - cross-domain-transfer/gaps.md: metaphor + narrative as
    cross-domain operators; narrative requires the Agency open-scope
    decision to pin first.
  - introspection/gaps.md: explain API is also the substrate for
    first-person narrative self-account.

Recommended v2 sequence in docs/PROGRESS.md:
  1. Pin Agency + Tool-use open-scope decisions (deadline: before
     Phase 3 engineering).
  2. Engineer Gaps 1 + 2 as one bounded PR.
  3. Engineer Gap 3 independently.
  4. Re-author cross-domain-transfer v2 with matched-control
     contract refinement.

Phase 3 v1 exit: 0/5 lanes passing, which is the expected v1 floor.
CLI suites smoke / cognition / teaching pass; no regression on
Phase 2.

2026-05-16 14:48:36 -07:00

2.2 KiB

Raw Blame History

multi-step-reasoning eval lane

What it measures

Whether the pipeline produces and consumes intermediate proposition-graph states for problems whose solution requires three or more inferential hops.

This sharpens inference-closure: inference-closure scored two-hop transitive entailments; this lane scores 3-, 4-, and 5-hop chains and additionally checks that intermediate states are observable in the proposition graph after the chain is taught.

Why it matters

Single-hop and two-hop closure can in principle be implemented by local pattern composition. Three-or-more hops require the pipeline to build and traverse an inference path that does not exist verbatim in any single premise. This is closer to the roadmap's question: does CORE think, or does it pattern-match longer templates.

Patterns covered (v1)

Pattern	Shape	Hops
`chain_3`	A is B; B is C; C is D	3
`chain_4`	A is B; B is C; C is D; D is E	4
`chain_5`	A is B; B is C; C is D; D is E; E is F	5
`mixed_relation_3`	A is B; B grounds C; C precedes D	3
`mixed_relation_4`	A causes B; B grounds C; C is D; D precedes E	4

Sub-metrics

M1. chain_endpoint_in_surface — the final-hop entity appears (case-insensitive, token-bounded) in surface or walk_surface.
M2. intermediate_in_graph — at least one intermediate hop is observable in the probe response's articulation_surface or walk_surface (proxy for graph state inspection).
M3. premises_stored — every taught hop emits a proposal.
M4. replay_determinism — two fresh runs match by trace_hash.

A case passes when M1 AND M3 AND M4 hold. M2 is reported as diagnostic signal — partial credit when chain_endpoint is missed.

Overall pass thresholds (v1)

chain_endpoint_recall_rate (M1) ≥ 0.50
premises_stored_rate ≥ 0.95
replay_determinism ≥ 0.95

Relationship to inference-closure v1

Same architectural gaps apply: no transitive composition in graph_planner.py, no path-recall in field/propagate.py. This lane scores how the gap scales with chain length. v1's likely result: uniform M1 failure across all chain lengths.

2.2 KiB Raw Blame History