Discussion, limitations, and implications
OPM v1.1.4 establishes a controlled behavioral transfer result, narrows several alternative explanations, and leaves the complete mechanism and efficiency program unresolved.
Continuous primitive sharing changed what the model could do outside the training matrix.
Every condition learned the observed combinations to near-ceiling validation and interpolation accuracy. The decisive difference appeared on familiar operations placed in familiar domains but combined in a way training never showed. Under that intervention, OPM_SHARED remained near ceiling while the domain-generalist control and both untied procedural controls degraded substantially.
This pattern is consistent with the central transfer theory: an operation implemented by parameters shared across domains remains trained and available when requested in a held-out domain. Domain-specific parameters do not receive that cross-domain experience. Because DOMAIN_GENERALIST received shared operation embeddings and learned every observed operation within each domain, the contrast is stronger than comparing OPM only with an entirely untrained operation copy.
Behavioral recombination advantage
A preregistered, sealed, seed-paired comparison supports H1 with a large interval-separated effect.
Primitive-path involvement
Large specific ablations and chance-level adapter-only performance suggest the primitive path carried necessary computation.
Complete causal reuse
The all-of criterion failed because one neural-probe family did not pass and two required inputs were unavailable.
Practical efficiency and autonomous discovery
Resource, hardware, exception-boundary, and learned-decomposition hypotheses require new preregistrations.
What the design rules out—and what it does not
The result is not readily explained by global training failure, because all conditions reached very high observed-cell validation and interpolation performance. It is not solely a comparison with an empty held-out module, because DOMAIN_GENERALIST had a trained transition for the target domain and a shared embedding for the target operation. It is also not evidence that oracle tokens directly encoded the label: the primitive path was restricted to selected facts and the adapter-only control returned performance to chance.
However, the experiment does not identify a unique internal ontology. Shared parameters may implement a computation that is functionally useful without corresponding one-to-one with a human-named relation. The study also cannot determine whether the same advantage survives learned parsing and routing, natural language, noisy bindings, broader task families, or domains with only partial semantic analogy.
8.3 · Frozen claim ledger
Every formal decision, including the negative ones.
| Claim identifier | Status | Observed evidence / disposition |
|---|---|---|
H1-PRIMARY | SUPPORTED | Δgeneralist = 0.426200; 95% CI [0.301938, 0.492272]; interpolation checks passed |
H2-THEORETICAL-RESOURCE-ADVANTAGE | NOT EVALUABLE | No preregistered quality-matched resource threshold |
H3-REAL-HARDWARE-ADVANTAGE | NOT EVALUABLE | Declared out of scope for v1 |
H4-CORRECT-BOUNDARY-OF-REUSE | NOT EVALUABLE | Declared out of scope for v1 |
MECHANISM-CAUSALLY-REUSABLE-PRIMITIVE | NOT SUPPORTED | Composite all-of criterion failed |
MET-005-INTERCHANGE-MECHANISM | NOT EVALUABLE | Accuracy 0.346271; no preregistered decision threshold |
MET-007-SURFACE-INVARIANCE | NOT EVALUABLE | Accuracy 0.999354; no preregistered decision threshold |
STA-004-ACTIVE-ABLATION | SUPPORTED | Every active operation met the ≥0.20 drop requirement in ≥2 trained domains |
STA-004-ADAPTER-ONLY | SUPPORTED | Maximum split accuracy 0.500125 ≤ 0.60 |
STA-004-NEURAL-PROBES | NOT SUPPORTED | OPM seed 4404, evidence step 1 failed Wilson/Holm criterion |
STA-004-RAW-ORACLE-PROBES | NOT EVALUABLE | Required aggregate-v4 input unavailable |
STA-004-RECOMBINATION | SUPPORTED | OPM_SHARED recombination accuracy 0.997811 > 0.80 |
STA-004-SENTINEL | SUPPORTED | Maximum logit change 0.0 < 1e-7 across 80 summaries |
STA-004-SYMBOLIC-ORACLE | NOT EVALUABLE | Required aggregate-v4 input unavailable |
STA-004-UNRELATED-PRESERVATION | SUPPORTED | Mean unrelated-operation drop was 0.0 for all four operations |
The result is deliberately narrow.
- Oracle decomposition.
Typed arguments, operation tokens, selected evidence, and execution order were supplied. The study did not test autonomous decomposition.
- Synthetic exact-isomorphism regime.
The three rendered domains were constructed to share the relevant formal operations. Natural domains may share only partial or contested structure.
- Deferred baselines.
Dense transformers, conventional MoE systems, oracle-informed MoE systems, and dense-execution primitive controls were outside the canonical v1 comparison.
- No quality-matched efficiency claim.
H2 lacked a preregistered resource threshold; H3 hardware endpoints were out of scope.
- No domain-exception test.
H4 requires a later partial-analogy regime with controlled exceptions and negative-transfer thresholds.
- Incomplete mechanism package.
One neural-probe family failed, and symbolic-oracle plus raw oracle-probe inputs were unavailable in aggregate v4.
Each open theory becomes a separate experiment.
These identifiers are prospective. They do not amend or reopen the closed v1.1.4 result.
OPM-MECH-COMPLETE-001Complete the mechanism evidence package and replicate neural probes.
OPM-MECH-INTERCHANGE-001Preregister context-aware interchange controls and numerical thresholds.
OPM-RESOURCE-001Test quality-matched parameter and FLOP advantages.
OPM-HARDWARE-001Measure latency, throughput, memory, and energy on declared hardware.
OPM-BOUNDARY-001Test domain-specific exceptions and negative-transfer boundaries.
Shared primitive organization improved controlled recombination generalization. The complete causal explanation remains an open scientific problem.