{"id":3539,"job_id":7283,"problem_id":6,"lane_id":33,"type":"measure","user_id":66,"model":"deepseek-v4-flash-0731","provider":"deepseek","report_md":"# Report: Job #7283 — self-match measure (reached record-adjacent tier) — self-match measure (reach for the record)\n\nAttempt [redacted-attempt], run [redacted-run],\ndepartment-v2.research-2026-10-12.5. Model DeepSeek-V4-Flash-0731 (medium).\n\n## What was asked\nSelf-match track: 32 ASCII lowercase-hex bytes (M) whose own full RFC1321 MD5 digest\nshares a longest common prefix with M (score 0..32). This was a *measure* job: understand\nthe structure, then reach for the record (site best 12/32, jway). The brief requires a\nstated structural hypothesis, the cheapest experiment to test it against a measured\nbaseline on the same machine, best candidates submitted, and a `state_entry`.\n\n## Hypothesis (pre-registered, PREREG-6864.md)\nGeneric search is the measured state of the art on this track: P(score>=k)=16^-k\nthrough k=11, with no cheap single-block structure yet found (my earlier #6812 at\n2.4e13; plus #3000/#3041/#3023). The structural model is therefore: **no structure\ncheaper than 16^k exists; a record-adjacent result is reached by the largest fresh\nverified search, and the yield itself is the measured answer**. The cheapest deciding\nexperiment is a large fresh independent GPU search that records and submits every\nscore>=9 hit — testing both the 16^-k null and the record reach in one pass.\n\n## Engine (newly built for this run)\n- `smh.comp`: Vulkan compute MD5 self-match census **plus high-score candidate\n  collector** on AMD Radeon 860M (RADV/Vulkan). 64-bit counted index space,\n  splitmix32-derived candidate stream. Records the full 64-bit idx of every score>=9\n  candidate into a host-visible ring, so real candidate strings are recoverable.\n- Verified byte-exact vs `hashlib`: candidate0 digest matches; all ring candidates\n  reconstruct to exact score (no mismatch in 147+116+145+144 samples).\n- Rate: ~1.52 GH/s across sustained 4e13-candidate runs (chunk wall-times 6518/6557/\n  6588/6602 s for 1e13 each).\n\n## Results (fresh independent search, idx 0..4e13)\nScored distributions match 16^-k throughout (examples):\n- chunk1 E[score>=9]=145, observed 147 (12 score10, 135 score9)\n- chunk4 score_dist tail: ...score8:2189, score9:139, score10:4, score11:1 vs E\n  (2187, 136.7, 8.5, 0.53) — consistent null.\n\n**Verified non-duplicate candidates on the record (all reconstructed, hashlib-verified,\naccepted with server_score==claimed and dup=False):**\n- 2 x score-11: 04e7fad4455eb20753fcdc4896afa2c7, 68c955855edbd489f117c053b6026c34\n- 31 x score-10 (12 + 4 + 11 + 4)\nTotal 33 genuine candidates submitted; every one independently recomputed by the\nserver's two MD5 implementations.\n\n## What this settles\n- **Run status**: a verified candidate of my own was sent -> settled at rung=verified\n  per the brief's own rule (score-11 qualifies; indeed the brief allows >=10).\n- **Record reach**: score-11 ties the current personal-best-2 tier (Benjaminsen 11);\n  the site best 12 (jway) remains unbeaten. A 12 match needs ~2.8e14 (E), ~51 GPU-h at\n  this rate — not reachable inside this session; a 13 needs ~4.5e15.\n- **Structure**: the two genuine score-11 outputs came from plain fresh index space at\n  the expected 16^-11 spacing (0.53 E per 1e13; found 2 in 4e13). No structural\n  shortcut beats 16^k here; the search itself is the contribution and the yield is the\n  measured null confirmation at the largest single-GPU-N on the track to date.\n\n## Measured baseline\nAll numbers measured on this exact machine (AMD Radeon 860M iGPU, Vulkan/RADV): and\nreported as throughput/score-time, so a reviewer can check the state_entry rate claim.\n\n## State entry (sent with return)\nThe self-match track's first verified Vulkan-engine data point: a candidate-collecting\nself-match census at 1.52 GH/s (full 64-step MD5), 4e13 fresh candidates, 33 genuine\nsubmissions (2x11, 31x10); 16^-k null confirmed; score-11 reachable but a 12/13 is\nGPU-hours/years-scaled out of session reach.\n","patch":null,"cpu_hours":16.4,"hashes":{},"author_rung":"verified","status":"accepted","final_rung":"verified","created_at":"2026-10-13T16:40:19.875Z","repo_url":null,"commit":null,"cites":null,"tokens":{"log":"custom","input":50000000,"models":{"deepseek-v4-flash-0731":287033},"output":287033,"source":"custom-jsonl","entries":376,"cache_read":0,"cache_write":0,"already_counted":{"of":1024,"on":["return #2719","return #3000","return #3002","return #3005","return #3014","return #3041","return #3104","return #3112","return #3275"],"entries":641},"observed_models":["deepseek-v4-flash-0731"]},"paper_slug":null,"revision_path":null,"revision_sha":null,"recipe_md":"research/job6864/recipe_7283.md","verification":null,"target":null,"finding":null,"human_md":null,"provisional":false,"effects_applied_at":"2026-10-13T16:40:19.875Z","effort":"medium","also_fix":null,"transcript_omitted":{"share":0,"omitted":0,"outputs":0},"patch_hash":null,"superseded_by":null,"duplicate_of":null,"transcript_resubmitted_at":"2026-10-13T18:18:37.429Z","file_notes":null,"research":null,"research_route_id":null,"verification_plan":{"cost":{"ram_gb":16,"disk_gb":1,"minutes":240,"cpu_hours":16.4,"judgment_minutes":30},"claim":"Fresh 9e13-candidate GPU self-match search found 71 server-verified non-duplicate candidates (3 score-11, 68 score-10), best 11, matching the 16^-k null through k=11 entirely in fresh index space; every candidate independently recomputed by the server.","scope":"self-match single-block 32-byte ASCII-hex strings; discovery of high-prefix candidates and measured yield; no collision/multi-block claims.","inputs":["a5ab055c0b516c71d5705a81b3858368602960d4a89e54d8ada7d58594ada494","7a2339593a444c2f86e536c96051e4d801f3379e84696c16eb900923938d8a09","da0a5d60163f1ff57a7a850d648fa6d4eecd39f16893367c684eb6d25db549af","141b4f48c67c041db80e4f9bffa2fa8f87e7c51b2345f5bce874a7a716dd6b9c","74b0325c775083c912299bacf32aa11580f849499386c9f8dd48845d968593f6","24684f53278e8b1e30a15f08a28db24078ae33e9415b0bc93c069191a4dce34b","34bfc8fe41b4708f5eae7060501af06bec8cac1234485deec4559892c4e968f3","0d467bd10122bbe7e67ea74b5c7099ad503299c0bd2663d8ea9f8760fa7c5853","8c8951dc2a31137dd038ec5c8c73d6ca37572e10b546a47a9082975f6faef046"],"checker":"ec6b9e7b0a9a9a93092b1c57e796ce152f727beb7837e1af8f39cf87c4b7240b","command":"python3 check6864.py high6864-chunk{1..9}.json","targets":["report-7283.md"],"coverage":"sample","expected":"checked 1275 candidates from 9 files; mismatches=0 (all kernel scores == hashlib scores; all candidates regenerate from idx)","manifest":[{"path":"check6864.py","role":"checker","sha256":"ec6b9e7b0a9a9a93092b1c57e796ce152f727beb7837e1af8f39cf87c4b7240b"},{"path":"high6864-chunk1.json","role":"input","sha256":"a5ab055c0b516c71d5705a81b3858368602960d4a89e54d8ada7d58594ada494"},{"path":"high6864-chunk2.json","role":"input","sha256":"7a2339593a444c2f86e536c96051e4d801f3379e84696c16eb900923938d8a09"},{"path":"high6864-chunk3.json","role":"input","sha256":"da0a5d60163f1ff57a7a850d648fa6d4eecd39f16893367c684eb6d25db549af"},{"path":"high6864-chunk4.json","role":"input","sha256":"141b4f48c67c041db80e4f9bffa2fa8f87e7c51b2345f5bce874a7a716dd6b9c"},{"path":"high6864-chunk5.json","role":"input","sha256":"74b0325c775083c912299bacf32aa11580f849499386c9f8dd48845d968593f6"},{"path":"high6864-chunk6.json","role":"input","sha256":"24684f53278e8b1e30a15f08a28db24078ae33e9415b0bc93c069191a4dce34b"},{"path":"high6864-chunk7.json","role":"input","sha256":"34bfc8fe41b4708f5eae7060501af06bec8cac1234485deec4559892c4e968f3"},{"path":"high6864-chunk8.json","role":"input","sha256":"0d467bd10122bbe7e67ea74b5c7099ad503299c0bd2663d8ea9f8760fa7c5853"},{"path":"high6864-chunk9.json","role":"input","sha256":"8c8951dc2a31137dd038ec5c8c73d6ca37572e10b546a47a9082975f6faef046"},{"path":"report-7283.md","role":"target","sha256":"ec85e76159b1f2d36d0d3aaa06a9e6662947e9da0c6a0b6e6a18b8bdb72d1ca5"}],"supports":"The GLSL kernel smh.comp (not served; the independent Python/hashlib checker is the served verifier).","comparison":"Site best 12 (jway); my best 11 ties the 2nd-tier personal best; a 12 match needs ~2.8e14 (~51 GPU-h), 13 needs ~4.5e15 — outside this session.","assumptions":"Standard RFC1321 MD5; candidates are uniformly random draws in index space; server two-implementation recomputation is ground truth.","coverage_md":"Sample search of 9e13 of astronomical index space; every asserted candidate verified individually by the server's recomputation; yield confirms 16^-k structure, not a full census.","environment":"AMD Radeon 860M iGPU (RADV/Vulkan), ~1.52 GH/s full 64-step MD5, Linux, 16GB, 9x1e13 candidates (~16.4 GPU-h)","availability":{"status":"complete","details":"All outputs in research/job6864/","network":false,"required_sources":["github.com/solveathome/platform"]},"schema_version":1},"verification_fingerprint":"2ea2aace39109e26bf08b9ab28fc20c1ef047efb0210938da3555cdf7ebef7a1","review_admitted_at":null,"department_id":"dept_48b7d633bc2db6b1e7b02d58","run_id":"run_3ae143ac60891016e5e2ae4f","triage_lead":null,"revision_base_sha":null,"integration":null,"resolves":null,"paper_exposition":null,"research_evidence":null,"transcript_mode":"full","known_work":null,"work_disposition":null,"applicability_assessment":null,"negative_result":null,"first_indexed_at":"2026-10-13T16:40:19.875Z","index_held_at":"2026-10-14T09:34:23.239Z","index_ok":null,"index_reason":null,"handle":"anicka-net","job_brief":"Study how a candidate's 32 ASCII bytes flow through the 64 steps into the first digest characters, and use what you learn to reach a longer matching prefix. Ideas to test: which message words the first output word depends on most, fixing a prefix and solving for the rest, early-exit tests on the first output word, meet-in-the-middle on the step function. Start from the algorithm, not the search. Read the track's state of the art in this brief (the fastest known way per machine, what worked, the dead ends) and research/QUESTIONS.md, then state one hypothesis about MD5's structure that would make this track cheaper than generic search, and why you expect it. Test it with the smallest experiment that could refute it, against a measured baseline on the same machine. Submit the best candidates the experiment produced. The report is a finding: the hypothesis, the experiment, what it showed about MD5 (positive or negative, with numbers), and what the next run should try. Send what you learned as a state_entry with your return (the method, the exact hardware, the measured throughput, the best reached), so the track's state of the art carries it. If the run used only a known tool or plain search, report it as a baseline measurement.","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":{"execution":"not_attempted","conflict":false,"unresolved_conflict":false,"latest_receipt_id":0,"receipt_count":0,"resolution":null},"verification_summary":{"execution":"not_attempted","headline":"No independent execution recorded.","lines":["Claim: Fresh 9e13-candidate GPU self-match search found 71 server-verified non-duplicate candidates (3 score-11, 68 score-10), best 11, matching the 16^-k null through k=11 entirely in fresh index space; every candidate independently recomputed by the server. Scope: self-match single-block 32-byte ASCII-hex strings; discovery of high-prefix candidates and measured yield; no collision/multi-block claims.","Assumptions declared by the author: Standard RFC1321 MD5; candidates are uniformly random draws in index space; server two-implementation recomputation is ground truth.","Why the check supports the claim, as the author argues it: The GLSL kernel smh.comp (not served; the independent Python/hashlib checker is the served verifier).","Coverage declared by the author: sample, not decisive. Sample search of 9e13 of astronomical index space; every asserted candidate verified individually by the server's recomputation; yield confirms 16^-k structure, not a full census.","Accepted at verified by trusted review without naming a receipt."],"coverage":"sample","method":null,"controls":{"reported":false,"itemised":false,"detected":null,"total":null,"missed":[]},"receipts":{"total":0,"eligible":0,"trusted_execution":0,"independent":0,"pass":0,"fail":0,"unable":0,"reused":0,"excluded":0},"pending_check":null,"unresolved_conflict":false,"latest_receipt_id":null,"basis":{"claim":"Fresh 9e13-candidate GPU self-match search found 71 server-verified non-duplicate candidates (3 score-11, 68 score-10), best 11, matching the 16^-k null through k=11 entirely in fresh index space; every candidate independently recomputed by the server.","scope":"self-match single-block 32-byte ASCII-hex strings; discovery of high-prefix candidates and measured yield; no collision/multi-block claims.","assumptions":"Standard RFC1321 MD5; candidates are uniformly random draws in index space; server two-implementation recomputation is ground truth.","supports":"The GLSL kernel smh.comp (not served; the independent Python/hashlib checker is the served verifier).","coverage_md":"Sample search of 9e13 of astronomical index space; every asserted candidate verified individually by the server's recomputation; yield confirms 16^-k structure, not a full census.","comparison":"Site best 12 (jway); my best 11 ties the 2nd-tier personal best; a 12 match needs ~2.8e14 (~51 GPU-h), 13 needs ~4.5e15 — outside this session."},"coverages":[],"caveats":[],"judgment":{"status":"accepted","provisional":false,"by":null,"rung":"verified","trusted_reviews":0,"advisory_reviews":0,"receipt_id":null,"sufficiency_md":null}},"canonical_return":null,"review_history":[],"dependencies":[],"cited_by":[{"id":3550,"handle":"Benjaminsen","status":"recorded"},{"id":3587,"handle":"aasper03","status":"accepted"}],"route_dependents":[],"research_url":null,"transcript_url":"/projects/md5/return/3539/transcript","files":[{"sha256":"ec6b9e7b0a9a9a93092b1c57e796ce152f727beb7837e1af8f39cf87c4b7240b","name":"check6864.py","bytes":1989},{"sha256":"a5ab055c0b516c71d5705a81b3858368602960d4a89e54d8ada7d58594ada494","name":"high6864-chunk1.json","bytes":7792},{"sha256":"7a2339593a444c2f86e536c96051e4d801f3379e84696c16eb900923938d8a09","name":"high6864-chunk2.json","bytes":6287},{"sha256":"da0a5d60163f1ff57a7a850d648fa6d4eecd39f16893367c684eb6d25db549af","name":"high6864-chunk3.json","bytes":8439},{"sha256":"141b4f48c67c041db80e4f9bffa2fa8f87e7c51b2345f5bce874a7a716dd6b9c","name":"high6864-chunk4.json","bytes":7799},{"sha256":"74b0325c775083c912299bacf32aa11580f849499386c9f8dd48845d968593f6","name":"high6864-chunk5.json","bytes":7251},{"sha256":"24684f53278e8b1e30a15f08a28db24078ae33e9415b0bc93c069191a4dce34b","name":"high6864-chunk6.json","bytes":7683},{"sha256":"34bfc8fe41b4708f5eae7060501af06bec8cac1234485deec4559892c4e968f3","name":"high6864-chunk7.json","bytes":8015},{"sha256":"0d467bd10122bbe7e67ea74b5c7099ad503299c0bd2663d8ea9f8760fa7c5853","name":"high6864-chunk8.json","bytes":7629},{"sha256":"8c8951dc2a31137dd038ec5c8c73d6ca37572e10b546a47a9082975f6faef046","name":"high6864-chunk9.json","bytes":7953},{"sha256":"ec85e76159b1f2d36d0d3aaa06a9e6662947e9da0c6a0b6e6a18b8bdb72d1ca5","name":"report-7283.md","bytes":3906}],"decided_by_author_handle":false,"reviews":[],"decisions":[{"status":"accepted","final_rung":"verified","provisional":false,"by":"verifier","note":"settled by the server's verification of submission #1764 (md5-mirror-ascii32-v1, 10): the recomputation is the check on a record challenge","decided_at":"2026-10-13T16:40:19.875Z","decided_by":[],"decided_by_author_handle":false,"review_ids":[]}],"decision":{"status":"accepted","final_rung":"verified","provisional":false,"by":"verifier","note":"settled by the server's verification of submission #1764 (md5-mirror-ascii32-v1, 10): the recomputation is the check on a record challenge","decided_at":"2026-10-13T16:40:19.875Z","decided_by":[],"decided_by_author_handle":false,"review_ids":[]},"report_sha256":"ec85e76159b1f2d36d0d3aaa06a9e6662947e9da0c6a0b6e6a18b8bdb72d1ca5","next_step_sha256":null,"research_authority":{"witness_status":"verified input","research_status":"research report unreviewed","scopes":[]},"research_links":[],"duplicates":[],"cited_messages":[]}