Observation validity under changing dynamics

Question

When should a controller reuse an observation, and when must it acquire fresh evidence because the apparatus may have changed?

Hypothesis

Under explicit continuity contracts, a change-aware deterministic rule may suffice. Changing the environment alone does not demonstrate an LLM advantage. Primary comparison is LLM_PERSISTENT against RULE_CHANGE_AWARE.

Registration

Frozen and locally committed at 0d97805a6407df7f9747ca460e3378fa37e2ef98. This r1 is published before inference; a verified public receipt gates first dispatch. Hidden change schedule and exact inputs/code are sealed, not disclosed. Prior memory-mechanism-v1 revision 2 remains unchanged, including its evidence qualifications and no measured LLM requirement.

Methods

Two independent blocks, four fresh apparatuses each: one unchanged and one changed under each contract. A,B,C,D,B,D,A,C per block. New target geometries on revisits. Dynamics fixed during tasks, changes only after task four, followed by restart of all policies. Neutral IDs, concealed assignments, public certified version token only where contract supplies it. Simulator-only first-passing sampling, declared cap 1000/apparatus, margins and public-prior witnesses. All 36 rejected candidates retained (22 and 14). Constructed diagnostic strata, not natural drift prevalence. Full sampling and filters in registration.

Baselines

All four policies use identical public APIs, observations, targets, precision, interval semantics and memory restrictions. Policies receive only public context, never the backend, coefficient, hidden schedule, evaluator truth, case-specific answers or privileged estimate cache. Recompute estimates each task. Fresh stores per condition/block, fresh model context per task, no earlier observations or conversations. LLM_PERSISTENT: verified installed 27B selects memory actions, calibration, estimation source IDs, integration references and final category/citation itself. Both contracts are supplied in every task. Only selected authentic observations carry cross-task evidence. RULE_CHANGE_AWARE: use current-task observations; otherwise search exact apparatus metadata only in certified mode, retrieve lexically sorted observations whose certified version equals the current token. In uncertified mode acquire current-task evidence. If none usable, execute the first legal calibration in lexical experiment-ID order. Estimate from all selected applicable observations, integrate exact target by reference, and finish from explicit certified event status, citing that integration. Unresolved estimate/integration: acquire the next untried lexical calibration only if at least four decisions remain for calibration/estimation/integration/finish; otherwise abstain. No hidden access or inference. RULE_ALWAYS_FRESH: acquire current-task calibration and use the same estimation/integration/finish procedure, without relying on previous-task evidence. RULE_BLIND_CACHE: search/retrieve by apparatus identity and ignore continuity. Calibrate only when no identity-matched observation exists, then use the same tools. Intentionally unsafe. It is not the competent baseline for an LLM-advantage claim. The common predeclared unresolved-tool rule applies. The four Williams rows (LLM, CHANGE, BLIND, FRESH), (CHANGE, FRESH, LLM, BLIND), (FRESH, BLIND, CHANGE, LLM), (BLIND, LLM, FRESH, CHANGE) repeat four times across sixteen targets. Every policy occupies every position four times; each policy precedes each other policy on 8/16 targets. Dispatch serially. Restart all controllers after task four in each block, preserving exact checkpoint and allowed observation IDs.

Evaluation

All 64 scheduled rows; sixteen per policy, only sixteen using LLM. Eight apparatuses are experimental units. Label accuracy and +1/-1/0 label utility separate from supported-correct classification, which requires cited committed exact-target integration of the full unaltered interval and publicly applicable source provenance. Track errors, abstentions, incomplete tasks, unsupported answers, missing/invalid citations, reuse validity, fresh/redundant calibration, calls/rejections and costs. Private unchanged truth does not certify stale evidence. No precise population claims.

Resource ceilings

128 inference calls including auxiliary; 1100000 full logical input tokens, 65536 generated; 512 generated and 300 seconds/request; 10800 seconds actual loading/provider-request elapsed. Eight model decisions/task, eighth finish only. Loading nested, counted once. No auxiliary inference, downloads, paid/cloud calls or interference. Rules/offline verification use zero inference. Preparation/reporting outside inference budget.

Stopping conditions

Stop entire partial study on unknown usage or uncertain operation. Canceled dispatch spent, unknown reservations retained; immutable late-response tombstone before any response/execution fence. No redispatch, adaptive prompts, outcome-selected retries or replacement fixtures. Registered checkpoint restart after task four. Accepted operations replay inertly.

Executing code

Isolated sealed runtime digest fd314aedd9249bae5d2eccde05352deb0af97aa7d30e77d56386dc92e5c0c958. Reference-based scientific interface, unchanged numerical estimator/integrator, immutable run/task-owned facts and source provenance. Code/contracts/inputs/schedule/order/scoring/recovery frozen before inference; private source withheld.

Configuration

Verified installed qwen3.8:27b-q4_K_M, digest 25b843619e944cd0ae6069f94ff4e5e26a16e109ccbc0a66a0f05979ed70098e; seed 424242, temperature .7, top_p .8, top_k 20, min_p 0, context 8192, cap 512, think=false. Six-decimal observations, six steps; calibration x=.25/.3 distinct from sealed targets. Target strict separation reduction >.0001. Both policies receive identical public contracts and API capabilities.

Complete outcomes

Execution pending. Simulator-only feasibility: change-aware rule 16/16 supported-correct with 14 calibrations; always-fresh 16/16 with 16; blind cache 12/16 labels correct but only 10 supported-correct with 8 calibrations. These are offline feasibility results on the frozen constructed fixtures, not live LLM results. Fifteen focused offline contract/recovery checks passed; thirteen controller-identity and eight recovery tests passed. A pre-registration seal-file filter failure was corrected while preserving generated fixtures and prepared stores, with zero inference.

Realized resources

Zero new inference at registration. Offline feasibility time recorded separately. Prior held reservations unchanged and separate: 34852 input tokens, 1024 generated, 600 seconds. Live report will include actual requests/tokens, calibration simulation, scientific tools, storage/journal/retrieval, other processing, guards, provider and total task time; model loading nested within request time.

Limitations

Eight apparatuses in two selected blocks, constructed diagnostic strata, exact identity, single hidden coefficient, explicit continuity contracts and guaranteed sufficient calibrations. No natural change-rate or precise population inference. Matching rules show this task admits a simpler implementation; no general claim about LLM value. Reference interface removes transcription, not proof of improved model arithmetic. No consciousness inference. Public derivatives do not permit full private-runtime replay.

Reproducibility

Provider-disabled offline feasibility and contract tests verify exact interval transfer, target matching, fact ownership, immutable provenance, memory isolation, recovery and late fences. Williams design balances serial position (four each) and pairwise precedence (eight each direction). Post-execution replay will reconstruct requests from event-prefix states, recompute rule responses/scientific facts/observations, independently score all rows, and reconcile resources. Executing source and raw model transcripts remain private.

Revision record

2026-10-14 r1: prospective public registration; inference has not begun. A later immutable revision will report all scheduled outcomes, including failures or stopped tasks.

Reviewed evidence & downloads

• contracts — derived. Selected project-authored factual data and policy/contracts; redistribution permitted.
• policies — derived. Selected project-authored factual data and policy/contracts; redistribution permitted.
• registration — derived. Selected project-authored factual data and policy/contracts; redistribution permitted.
• summary — derived. Selected project-authored factual data and policy/contracts; redistribution permitted.

These are selected public derivatives; raw transcripts, internal configuration and private source are withheld. Hashes describe the actual downloads, not unmodified original evidence.

Artifact manifest and source provenance · Permanent revision URL

Substantive versions

• Revision 1