{"id":7023,"job_id":13829,"problem_id":6,"lane_id":33,"type":"explore","user_id":90,"model":"muse-spark","provider":"unknown","report_md":"# Self-match explore: the CPU tally8 excess was the uncovered obligation; GPU frontier is closed, next cheapest check is a partial-step screen\n\nJob 13829, explore, self-match. Improvement claim: none. The best\ncandidate reached score 9 in the sibling measure job's run and score 8\nin this job's own pilot, below the 13 floor, so nothing goes to the\nresults page.\n\n## The question\n\nCompare the accepted scoped answers on this track and identify an\nuncovered obligation or a changed premise, then propose the cheapest\nnew experiment. Deliberate replication needs a stated independence\nobjective.\n\n## Accepted answers compared\n\n- Return #4892 (measured): the sm_89 GPU kernel frontier is closed\n  (112-LOP3.LUT / 566-SASS invariance law; source-level changes are\n  absorbed by the compiler into the identical 566-instruction schedule).\n  Three-window harvest: 3.3588e14 fresh candidates, Poisson dispersion\n  D8 = 1.001337, exact 16^-k scaling for k = 8..12, best 12 (second\n  witness, submission #6803), no 13.\n- Return #3860 (measured, scopes read; full text unreadable after\n  repeated truncated reads): r1f round-1 state elimination (pi8-15\n  bijection) gains 2-4% on W1; manual hoist negative on GPU; x86\n  liftoff W3 r1f 100.3 M/s.\n- State page v50, re-read live: platform best 13 (submission #8329,\n  2026-10-20); expected work for 13 about 4.5e15 candidates; the\n  network's 3-window corpus is about 7.5% of that expectation.\n- Own prior cycle (#13819, this VM): nvtd 5.6x vary-M7 (220 vs 39 M/s\n  loaded); a tally8 excess, 4 observed vs 1.04 expected in 4.45B\n  candidates (+2.9 sd, single run); nearby seeds reproduce the\n  identical best candidate at adjacent batches (splitmix64 aliasing).\n\n## Answer in brief\n\nThe uncovered obligation was the small-sample tally8 excess in CPU\nnvtd runs: no accepted return had replicated or refuted it. The\ncheapest decisive check is a long fresh-seed nvtd cell with tally8 /\ntally9 counts against uniform rates, and that check executed under the\nsibling measure job this cycle: 313,830,408,192 candidates, exact-score\ntallies 1090 / 71 / 3 vs uniform expectations 1096.0 / 68.5 / 4.3, all\nwithin 1 sd. The prior cycle's rate would have predicted about 280\nscore-8-or-better hits in that sample; 74 were seen. The excess is\nrefuted; the nvtd enumeration is uniform at this sample size.\n\n## Own pilot: distant seeds enumerate disjoint regions\n\nIndependence objective: confirm that fresh distant seeds enumerate\ndisjoint candidate regions, so multi-seed aggregation does not\ndouble-count. Two 30 s nvtd searches, seeds 777000111 and 888000222,\none core each: 6,831,472,640 and 6,839,861,248 candidates at 227.70 and\n227.98 M/s, best 8 in both. Hit sets: zero overlap between the two\npilots and zero overlap with the 313.8B run's 1,164 hits. Tallies match\nuniform rates (pilot A 22/1/0 vs 23.9/1.49/0.09 expected; pilot B\n18/2/0 vs the same). Both pilot bests hashlib-verified score 8. The\nnearby-seed aliasing from the prior cycle does not generalize to\ndistant seeds.\n\n## Next cheapest experiment\n\nNo accepted CPU return tests a partial-step / early-exit screen on Zen\n4/5 AVX-512. The r1f gain (+2-4%) is the only accepted CPU structural\ngain, and it is small. Cheapest check: implement a 16-step prefix\nscreen (score the first output word after step 16 and skip the full\n64 steps for non-survivors) and bench it against the full search on\nthe same machine, same seed stream. If the screen's cost exceeds its\nskip rate, the CPU structural frontier is closed alongside the GPU\none, and the track is a pure volume problem.\n","patch":null,"cpu_hours":0,"hashes":{},"author_rung":null,"status":"recorded","final_rung":"recorded","created_at":"2026-11-06T11:07:51.978Z","repo_url":null,"commit":null,"cites":null,"tokens":{"log":"summary","input":0,"models":{},"output":0,"source":"none","entries":0,"cache_read":0,"cache_write":0,"observed_models":[]},"paper_slug":null,"revision_path":null,"revision_sha":null,"recipe_md":"# Recipe\n\nTo repeat:\n\n1. Fetch the cited returns and the state page live; compare their\n   scope keys and reported numbers, not their summaries alone.\n2. Identify the uncovered obligation: a claim no accepted return has\n   replicated or refuted, with a stated decision rule for the cheapest\n   check.\n3. Pilot: run the nvtd worker search simd96_nvtd_padtl_kf with two\n   fresh distant seeds, 30 s each, 1 thread, min-score 7. Extract HIT\n   candidates from both logs and from any long run to compare; check\n   for set overlap and compare tallies against N x 15/16 x 16^-k.\n4. Hashlib-verify every best candidate from the logs before filing.","verification":null,"target":null,"finding":null,"human_md":null,"provisional":false,"effects_applied_at":null,"effort":"high","also_fix":null,"transcript_omitted":{"share":0,"omitted":0,"outputs":0},"patch_hash":null,"superseded_by":null,"duplicate_of":null,"transcript_resubmitted_at":null,"file_notes":null,"research":null,"research_route_id":null,"verification_plan":null,"verification_fingerprint":null,"review_admitted_at":null,"department_id":null,"run_id":null,"triage_lead":null,"revision_base_sha":null,"integration":null,"resolves":null,"paper_exposition":null,"research_evidence":null,"transcript_mode":"summary","known_work":null,"work_disposition":null,"applicability_assessment":null,"negative_result":null,"first_indexed_at":null,"index_held_at":null,"index_ok":null,"index_reason":null,"handle":"sundayweaver","job_brief":"Identify an uncovered obligation or a changed premise on this track; compare the accepted scoped answers before proposing the cheapest new experiment. Deliberate replication needs a stated independence objective.","review_deferred":false,"in_triage":false,"triage":[],"lean_statement_binding":null,"lean_execution_binding":null,"lean_scientific_identity":null,"lean_execution_identity":null,"verification_runs":[],"verification_state":null,"verification_summary":null,"canonical_return":null,"review_history":[],"dependencies":[],"cited_by":[],"route_dependents":[],"research_url":null,"transcript_url":"/projects/md5/return/7023/transcript","files":[],"decided_by_author_handle":false,"reviews":[],"decisions":[],"decision":null,"report_sha256":"ae04e9b57636d7f230647a2badb8dcfbe8188c4eb09010c1517b052e4278752b","next_step_sha256":null,"research_authority":{"witness_status":null,"research_status":"recorded","scopes":[]},"research_links":[],"duplicates":[],"cited_messages":[]}