[ 0.0 s] ==================================================================================================== [ 0.0 s] Stage 2, decisive (registered), started 2026-11-08T01:21:12Z; PREREGISTRATION.md with Amendments 1-15; argv --decisive --workers 4 [ 0.0 s] HEAD 6b6ec0996342ae4befdab93b3f2179fea0cc9f8c; the tree is clean [ 0.0 s] load average [0.01, 0.02, 0.24] (start) [ 0.0 s] interpreter flags: isolated (-I) True, no bytecode (-B) True [ 0.0 s] environment b44f374b1412: Python 3.13.18, numpy 2.5.3, numba 0.68.1 [ 0.0 s] the instance manifest is frozen by Amendment 14 [ 82.3 s] count checks (word-level, A13.9): all equal (81.3 s) [ 85.8 s] 32 instances: BD 12365, BD-at 12334, BD-dir 12365, BD-dir-at 12334, BD-dir-glove 12339, BD-glove 12339, BD-suite 13729, BIB-all 1658, EVAL-test 5382, HIT-test 29383, HIT-val 29383, HL 2163, HL-at 2153, HL-glove 2162, HL-split 541, LEDS-test 2495, MEN 2005, MEN-glove 2005, SHW-test 17073, SL 666, SL-glove 666, WB 1493, WB-all 1658, WB-at 1491, WB-glove 1493, WB-suite 1658, WSC 344, WSC-glove 344, WSR 245, WSR-glove 245, WSS 197, WSS-glove 197 [ 85.8 s] wiring: DECISION_SPEC (69 comparisons) and CONTROL_SPEC (10 controls) equal Amendment 13's text [ 85.9 s] pre-flight: tuning at b2ba411c3ba8, changed since: Pages/Hyperbolic Lexicon/PREREGISTRATION.md, Pages/Hyperbolic Lexicon/reference/results/stage2-tuning-console.txt, Pages/Hyperbolic Lexicon/reference/results/tuning-stage2.json, Pages/Hyperbolic Lexicon/reference/settings-stage2.json; the lr freeze is Amendment 15 [ 106.9 s] the instance manifest rebuilt from the pinned files equals the committed one; the pair sets are its rows [ 111.5 s] disk pre-flight (the decisive run): 1.363 GB to write, 12.91 GB free; within the free space less 2 GB [ 111.5 s] pass 1, exact arms: 453 jobs (1128 code files) [ 111.5 s] pass 1: exact arms: 453 jobs, 0 done before (runs.jsonl), 453 to run, 4 worker process(es) (fork) [ 111.5 s] load average [0.84, 0.33, 0.33] (pass 1: exact arms: start) [ 118.4 s] [1/453] p1-H3-BD-dir (6.8 s; 7 s in) [ 124.7 s] [2/453] p1-F2-HL (6.2 s; 13 s in) [ 131.5 s] [3/453] p1-H3-HIT-test (19.9 s; 20 s in) [ 131.8 s] [4/453] p1-F2s-HL (7.1 s; 20 s in) [ 132.0 s] [5/453] p1-H3-HL (20.4 s; 20 s in) [ 132.3 s] [6/453] p1-H2-HIT-test (20.7 s; 21 s in) [ 137.5 s] [7/453] p1-F3-HL (6.0 s; 26 s in) [ 139.2 s] [8/453] p1-F3s-HL (7.4 s; 28 s in) [ 151.4 s] [9/453] p1-H2-HL (19.5 s; 40 s in) [ 154.5 s] [10/453] p1-K3-HL (22.2 s; 43 s in) [ 155.8 s] [11/453] p1-S-full-0-HL (18.2 s; 44 s in) [ 158.1 s] [12/453] p1-S-full-1-HL (18.9 s; 47 s in) [ 170.6 s] [13/453] p1-S-full-10-HL (19.2 s; 59 s in) [ 173.7 s] [14/453] p1-S-full-11-HL (19.3 s; 62 s in) [ 174.1 s] [15/453] p1-S-full-12-HL (18.4 s; 63 s in) [ 176.7 s] [16/453] p1-S-full-13-HL (18.6 s; 65 s in) [ 189.4 s] [17/453] p1-S-full-14-HL (18.8 s; 78 s in) [ 191.6 s] [18/453] p1-S-full-15-HL (17.9 s; 80 s in) [ 192.6 s] [19/453] p1-S-full-16-HL (18.5 s; 81 s in) [ 194.4 s] [20/453] p1-S-full-17-HL (17.7 s; 83 s in) [ 226.4 s] [25/453] p1-S-full-4-HL (18.4 s; 115 s in) [ 328.4 s] [50/453] p1-F2s-BD-dir (9.2 s; 217 s in) [ 361.5 s] [75/453] p1-S-full-8-BD-dir (4.9 s; 250 s in) [ 391.7 s] [100/453] p1-F2-SHW-test (6.8 s; 280 s in) [ 412.2 s] [125/453] p1-K3-BD (7.0 s; 301 s in) [ 427.3 s] [150/453] p1-W-16-BD (3.1 s; 316 s in) [ 442.3 s] [175/453] p1-H3-BIB-all (0.5 s; 331 s in) [ 451.4 s] [200/453] p1-GloVe-vec-16-HL-glove (0.1 s; 340 s in) [ 459.4 s] [225/453] p1-GloVe-vec-13-MEN-glove (0.1 s; 348 s in) [ 460.7 s] [250/453] p1-S-full-14-MEN (0.1 s; 349 s in) [ 462.0 s] [275/453] p1-W-19-MEN (0.1 s; 350 s in) [ 464.5 s] [300/453] p1-F2s-WB (0.5 s; 353 s in) [ 468.0 s] [325/453] p1-W-0-WB (0.2 s; 356 s in) [ 470.2 s] [350/453] p1-F3-SL (0.0 s; 359 s in) [ 471.3 s] [375/453] p1-S-full-8-SL (0.0 s; 360 s in) [ 472.2 s] [400/453] p1-F2s-HL-split (0.3 s; 361 s in) [ 478.5 s] [425/453] p1-GloVe-vec-13-WSR-glove (0.0 s; 367 s in) [ 478.8 s] [450/453] p1-H3-WSS (0.0 s; 367 s in) [ 482.2 s] [453/453] p1-GloVe-vec-1-WSR-glove (6.1 s; 371 s in) [ 482.2 s] load average [4.0, 2.95, 1.56] (pass 1: exact arms: end) [ 483.7 s] training: measured, edge-shuffled, pre-flight, replay: 43 jobs, 0 done before (runs.jsonl), 43 to run, 4 worker process(es) (spawn) [ 483.7 s] load average [3.76, 2.92, 1.55] (training: measured, edge-shuffled, pre-flight, replay: start) [ 1083.7 s] load average [4.0, 3.88, 2.74] (training: measured, edge-shuffled, pre-flight, replay) [ 1683.7 s] load average [4.0, 4.0, 3.37] (training: measured, edge-shuffled, pre-flight, replay) [ 1946.4 s] [1/43] measured-lorentz-d3-lr10-s2 (1459.2 s; 1463 s in) [ 1948.9 s] [2/43] measured-lorentz-d3-lr10-s1 (1461.8 s; 1465 s in) [ 1971.1 s] [3/43] measured-lorentz-d3-lr10-s0 (1484.2 s; 1487 s in) [ 2331.1 s] load average [4.04, 4.01, 3.71] (training: measured, edge-shuffled, pre-flight, replay) [ 2931.1 s] load average [4.0, 4.01, 3.88] (training: measured, edge-shuffled, pre-flight, replay) [ 3405.6 s] [4/43] measured-lorentz-d3-lr10-s4 (1456.6 s; 2922 s in) [ 3408.9 s] [5/43] measured-lorentz-d3-lr10-s3 (1462.5 s; 2925 s in) [ 3458.9 s] [6/43] shuffled-lorentz-d3-lr10-s0 (1487.8 s; 2975 s in) [ 3578.9 s] load average [4.0, 4.01, 3.97] (training: measured, edge-shuffled, pre-flight, replay) [ 3680.0 s] [7/43] replay-measured-lorentz-d3-lr10-s0 (3193.1 s; 3196 s in) [ 4220.0 s] load average [4.02, 4.01, 4.0] (training: measured, edge-shuffled, pre-flight, replay) [ 4820.0 s] load average [4.0, 4.0, 4.0] (training: measured, edge-shuffled, pre-flight, replay) [ 4949.1 s] [8/43] shuffled-lorentz-d3-lr10-s2 (1540.2 s; 4465 s in) [ 4958.0 s] [9/43] shuffled-lorentz-d3-lr10-s1 (1552.5 s; 4474 s in) [ 5015.2 s] [10/43] shuffled-lorentz-d3-lr10-s3 (1556.3 s; 4532 s in) [ 5234.2 s] [11/43] shuffled-lorentz-d3-lr10-s4 (1554.2 s; 4750 s in) [ 5474.2 s] load average [4.0, 4.0, 4.0] (training: measured, edge-shuffled, pre-flight, replay) [ 5943.0 s] [12/43] measured-euclid-d3-lr10-s0 (993.9 s; 5459 s in) [ 5969.7 s] [13/43] measured-euclid-d3-lr10-s1 (1011.7 s; 5486 s in) [ 6024.3 s] [14/43] measured-euclid-d3-lr10-s2 (1009.1 s; 5541 s in) [ 6084.3 s] load average [4.04, 4.01, 4.0] (training: measured, edge-shuffled, pre-flight, replay) [ 6224.0 s] [15/43] measured-euclid-d3-lr10-s3 (989.8 s; 5740 s in) [ 6704.0 s] load average [4.0, 4.0, 4.0] (training: measured, edge-shuffled, pre-flight, replay) [ 6945.4 s] [16/43] measured-euclid-d3-lr10-s4 (1002.4 s; 6462 s in) [ 6993.4 s] [17/43] shuffled-euclid-d3-lr10-s0 (1023.6 s; 6510 s in) [ 7049.0 s] [18/43] shuffled-euclid-d3-lr10-s1 (1024.6 s; 6565 s in) [ 7245.9 s] [19/43] shuffled-euclid-d3-lr10-s2 (1021.9 s; 6762 s in) [ 7305.9 s] load average [4.0, 4.0, 4.0] (training: measured, edge-shuffled, pre-flight, replay) [ 7905.9 s] load average [4.0, 4.0, 4.0] (training: measured, edge-shuffled, pre-flight, replay) [ 7970.6 s] [20/43] shuffled-euclid-d3-lr10-s3 (1025.1 s; 7487 s in) [ 8518.5 s] load average [4.0, 4.0, 4.0] (training: measured, edge-shuffled, pre-flight, replay) [ 9133.9 s] load average [4.0, 4.01, 4.0] (training: measured, edge-shuffled, pre-flight, replay) [ 9330.4 s] [25/43] measured-lorentz-d2-lr10-s3 (1303.9 s; 8847 s in) [ 9757.5 s] load average [4.0, 4.0, 4.0] (training: measured, edge-shuffled, pre-flight, replay) [ 10373.4 s] load average [4.0, 4.0, 4.0] (training: measured, edge-shuffled, pre-flight, replay) [ 11011.0 s] load average [4.0, 4.0, 4.0] (training: measured, edge-shuffled, pre-flight, replay) [ 11619.2 s] load average [4.01, 4.0, 4.0] (training: measured, edge-shuffled, pre-flight, replay) [ 12235.4 s] load average [4.0, 4.0, 4.0] (training: measured, edge-shuffled, pre-flight, replay) [ 12879.6 s] load average [4.0, 4.0, 4.0] (training: measured, edge-shuffled, pre-flight, replay) [ 13484.2 s] load average [2.08, 2.4, 3.16] (training: measured, edge-shuffled, pre-flight, replay) [ 13754.5 s] [43/43] shuffled-euclid-d2-lr10-s4 (828.3 s; 13271 s in) [ 13755.3 s] load average [1.02, 1.62, 2.64] (training: measured, edge-shuffled, pre-flight, replay: end) [ 13755.3 s] replay check: measured-lorentz-d3-lr10-s0 after epoch 1000: EQUAL [ 13768.6 s] floors: 20 untrained tables (Rng(2600 + i), i < 5, every cell), snapped [ 13768.7 s] pass 1: learned tables: 108 jobs, 0 done before (runs.jsonl), 108 to run, 4 worker process(es) (fork) [ 13768.7 s] load average [1.02, 1.6, 2.62] (pass 1: learned tables: start) [ 13939.6 s] [1/108] p1L-decisive-measured-euclid-d2-lr10-s3 (170.9 s; 171 s in) [ 13940.4 s] [2/108] p1L-decisive-measured-euclid-d2-lr10-s1 (171.6 s; 172 s in) [ 13940.7 s] [3/108] p1L-decisive-measured-euclid-d2-lr10-s2 (171.9 s; 172 s in) [ 13941.2 s] [4/108] p1L-decisive-measured-euclid-d2-lr10-s0 (172.5 s; 173 s in) [ 14108.7 s] [5/108] p1L-decisive-measured-euclid-d2-lr10-s4 (169.1 s; 340 s in) [ 14111.6 s] [6/108] p1L-decisive-measured-euclid-d3-lr10-s0 (171.2 s; 343 s in) [ 14112.2 s] [7/108] p1L-decisive-measured-euclid-d3-lr10-s2 (170.9 s; 343 s in) [ 14112.4 s] [8/108] p1L-decisive-measured-euclid-d3-lr10-s1 (171.8 s; 344 s in) [ 14280.4 s] [9/108] p1L-decisive-measured-euclid-d3-lr10-s3 (171.7 s; 512 s in) [ 14285.1 s] [10/108] p1L-decisive-measured-euclid-d3-lr10-s4 (173.5 s; 516 s in) [ 14366.5 s] [11/108] p1L-decisive-measured-lorentz-d2-lr10-s1 (254.1 s; 598 s in) [ 14368.0 s] [12/108] p1L-decisive-measured-lorentz-d2-lr10-s0 (255.8 s; 599 s in) [ 14428.0 s] load average [4.0, 3.75, 3.35] (pass 1: learned tables) [ 14534.9 s] [13/108] p1L-decisive-measured-lorentz-d2-lr10-s2 (254.5 s; 766 s in) [ 14539.1 s] [14/108] p1L-decisive-measured-lorentz-d2-lr10-s3 (254.0 s; 770 s in) [ 14621.8 s] [15/108] p1L-decisive-measured-lorentz-d2-lr10-s4 (255.3 s; 853 s in) [ 14624.7 s] [16/108] p1L-decisive-measured-lorentz-d3-lr10-s0 (256.7 s; 856 s in) [ 14791.9 s] [17/108] p1L-decisive-measured-lorentz-d3-lr10-s1 (257.0 s; 1023 s in) [ 14796.8 s] [18/108] p1L-decisive-measured-lorentz-d3-lr10-s2 (257.7 s; 1028 s in) [ 14832.9 s] [19/108] p1L-decisive-shuffled-euclid-d2-lr10-s0 (41.0 s; 1064 s in) [ 14837.9 s] [20/108] p1L-decisive-shuffled-euclid-d2-lr10-s1 (41.1 s; 1069 s in) [ 14917.2 s] [25/108] p1L-decisive-shuffled-euclid-d2-lr10-s4 (42.4 s; 1148 s in) [ 15032.8 s] load average [4.0, 3.98, 3.68] (pass 1: learned tables) [ 15104.7 s] [50/108] p1L-decisive-floor-lorentz-d2-i0 (22.0 s; 1336 s in) [ 15278.8 s] [75/108] p1L-decisive-tuning-euclid-d3-lr0.1-s9002 (32.9 s; 1510 s in) [ 15462.5 s] [100/108] p1L-decisive-tuning-lorentz-d3-lr0.1-s9003 (29.7 s; 1694 s in) [ 15522.3 s] [108/108] p1L-decisive-tuning-lorentz-d3-lr3-s9009 (29.7 s; 1754 s in) [ 15522.3 s] load average [3.76, 3.95, 3.81] (pass 1: learned tables: end) [ 15523.7 s] count checks (labels and TYPE, pass 2): all equal [ 15523.7 s] HyperLex random split (A13.11): the split file and hyperlex_rnd.tsv agree (words, TYPE, AVG_SCORE; 541 noun rows, 541 covered) [ 15552.1 s] L12(a): supported [ 15552.1 s] L12(b): supported [ 15552.1 s] L12(c): supported [ 15552.1 s] L12(d): refuted [ 15552.1 s] L12 controls fail as required: {'control1': True, 'control2': True, 'control3': True, 'control4': True} [ 15594.0 s] L13(a): supported [ 15594.0 s] L13(b): supported [ 15594.0 s] L13(c): supported [ 15594.0 s] L13 controls fail as required: {'control1': True, 'control2': True, 'control3': True} [ 15817.9 s] L14 closure check (must pass): {'random': True, 'hard': True} [ 15817.9 s] L14(a): supported [ 15817.9 s] L14(b): supported [ 15817.9 s] L14 controls 1-2 and the closure check fail/pass as required: {'control1': True, 'control2': True, 'closure_check': True, 'control3': True} [ 15849.3 s] L15(a): refuted [ 15849.3 s] L15(b): supported [ 15849.3 s] L15 controls fail as required: {'control1': True, 'control2': True, 'control3': True} [ 16140.3 s] wiring check: 141 decided comparisons and 321 control values equal A13's text and their recomputation from the files [ 16681.9 s] wrote results/claims-stage2.json (SHA-256 b6e4a75b01010248cb5e6f431bc9ce63a656e342ed1fd05376a2301a15e46698), results/claims-stage2.md, results/lbench-stage2.txt (1848 records), results/stage2-manifest.json; decision code files hashes only (68.8 MB), measured snapped texts committed (26.2 MB) [ 16683.9 s] load average [1.0, 1.07, 1.79] (end) [ 16683.9 s] decisive finished with exit code 0 (16684 s)