{ "what": "Key LM bits-per-byte contrasts cited in README.md (hierarchical cluster bootstrap, 10,000 replicates). Full tables: dev_lm_confirm_ranking.csv, test_lm_confirm.csv, dev_lm_large.csv, test_lm_large.csv.", "dev_large_top_set_pairs": { "R2-A4-SPnat-D2-32k - R2-A10-MinGram-P1r3-D2-48k": { "delta": -0.010331, "delta_pct": -0.903037, "ci95": [ -0.015725, -0.006244 ], "ci95_pct": [ -1.347158, -0.557606 ], "ci90": [ -0.014546, -0.006893 ], "p": 0.0, "p_plus1": 0.0002, "se": 0.002375, "se_pct": 0.200841, "tost_equivalent": false, "ci95_excludes_0": true, "mde80_pct": 0.562673 }, "R2-A4-SPnat-D2-32k - R2-A10-MinGram-P1r3-D2-32k": { "delta": -0.00713, "delta_pct": -0.625, "ci95": [ -0.012026, -0.003405 ], "ci95_pct": [ -1.043154, -0.305626 ], "ci90": [ -0.01119, -0.003914 ], "p": 0.0004, "p_plus1": 0.0006, "se": 0.00221, "se_pct": 0.18929, "tost_equivalent": false, "ci95_excludes_0": true, "mde80_pct": 0.530313 }, "R2-A10-MinGram-P1r3-D2-32k - R2-A10-MinGram-P1r3-D2-48k": { "delta": -0.003201, "delta_pct": -0.279786, "ci95": [ -0.005269, -0.000834 ], "ci95_pct": [ -0.46409, -0.073777 ], "ci90": [ -0.004942, -0.001272 ], "p": 0.011, "p_plus1": 0.011199, "se": 0.001122, "se_pct": 0.098256, "tost_equivalent": false, "ci95_excludes_0": true, "mde80_pct": 0.275272 } }, "test_large_boot_2seeds_3cands": { "R2-A4-SPnat-D2-32k - A1-P1r3-D2-16k": { "delta": -0.008684786018473334, "delta_pct": -0.7701554030632566, "ci95_pct": [ -1.1521793550485442, -0.3981661969961223 ], "ci90": [ -0.012104198271208922, -0.005240198401928611 ], "p": 0.0, "p_plus1": 0.00019998000199980003, "se_pct": 0.19094983287369818, "ci95_excludes_0": true, "tost_equivalent": false, "mde80_pct": 0.5349622291800946 }, "R2-A10-MinGram-P1r3-D2-48k - A1-P1r3-D2-16k": { "delta": -0.0028933137990660995, "delta_pct": -0.25657526280652554, "ci95_pct": [ -0.4354746500167368, -0.08159379637904507 ], "ci90": [ -0.004526817078030976, -0.0012422862460522944 ], "p": 0.004, "p_plus1": 0.004199580041995801, "se_pct": 0.09050337810749172, "ci95_excludes_0": true, "tost_equivalent": false, "mde80_pct": 0.2535529262952376 }, "R2-A4-SPnat-D2-32k - R2-A10-MinGram-P1r3-D2-48k": { "delta": -0.005791472219407234, "delta_pct": -0.5149012494908062, "ci95_pct": [ -0.8634677179864138, -0.21702516982561376 ], "ci90": [ -0.008881838988751922, -0.0029836125164287245 ], "p": 0.0014, "p_plus1": 0.0015998400159984002, "se_pct": 0.16320821808817637, "ci95_excludes_0": true, "tost_equivalent": false, "mde80_pct": 0.4572417312703925 } }, "test_large_boot_seed1_4cands": { "R2-A4-SPnat-D2-32k - A1-P1r3-D2-16k": { "delta": -0.009749268125608612, "delta_pct": -0.8640569198004335, "ci95_pct": [ -1.2490589169222628, -0.5221700105007353 ], "ci90": [ -0.013105525700701049, -0.006637920011845844 ], "p": 0.0, "p_plus1": 0.00019998000199980003, "se_pct": 0.18159743082505547, "ci95_excludes_0": true, "tost_equivalent": false, "mde80_pct": 0.5087606778467678 }, "R2-A10-MinGram-P1r3-D2-32k - A1-P1r3-D2-16k": { "delta": -0.005916025958605076, "delta_pct": -0.5243248109901171, "ci95_pct": [ -0.7333558799186555, -0.2505256598055685 ], "ci90": [ -0.0078102004497957785, -0.003451502718269095 ], "p": 0.0002, "p_plus1": 0.00039996000399960006, "se_pct": 0.12239615123434773, "ci95_excludes_0": true, "tost_equivalent": false, "mde80_pct": 0.34290324805206795 }, "R2-A10-MinGram-P1r3-D2-48k - A1-P1r3-D2-16k": { "delta": -0.0035672107921167484, "delta_pct": -0.31615431328829585, "ci95_pct": [ -0.4909429266021485, -0.16054069054834635 ], "ci90": [ -0.005149563958236636, -0.0020913877814648712 ], "p": 0.0, "p_plus1": 0.00019998000199980003, "se_pct": 0.08528340828868017, "ci95_excludes_0": true, "tost_equivalent": false, "mde80_pct": 0.23892873601185938 }, "R2-A4-SPnat-D2-32k - R2-A10-MinGram-P1r3-D2-32k": { "delta": -0.0038332421670035366, "delta_pct": -0.34152279757318016, "ci95_pct": [ -0.7391926839990545, -0.10171298736361388 ], "ci90": [ -0.00736603239301361, -0.0014457211118628212 ], "p": 0.0, "p_plus1": 0.00019998000199980003, "se_pct": 0.162594645953095, "ci95_excludes_0": true, "tost_equivalent": false, "mde80_pct": 0.45552275664650255 }, "R2-A4-SPnat-D2-32k - R2-A10-MinGram-P1r3-D2-48k": { "delta": -0.006182057333491864, "delta_pct": -0.549640318085336, "ci95_pct": [ -0.8940447142057342, -0.2488457905349242 ], "ci90": [ -0.00924963772336348, -0.0033845498604445205 ], "p": 0.0006, "p_plus1": 0.0007999200079992001, "se_pct": 0.16206454279153712, "ci95_excludes_0": true, "tost_equivalent": false, "mde80_pct": 0.4540376274650071 }, "R2-A10-MinGram-P1r3-D2-48k - R2-A10-MinGram-P1r3-D2-32k": { "delta": 0.0023488151664883272, "delta_pct": 0.20926774038606374, "ci95_pct": [ -0.06444365120164039, 0.44287168269519417 ], "ci90": [ -0.0002884015121531357, 0.004546200404592969 ], "p": 0.1366, "p_plus1": 0.1367863213678632, "se_pct": 0.12949901051134746, "ci95_excludes_0": false, "tost_equivalent": false, "mde80_pct": 0.36280251360884697 } }, "test_large_parametric_all_seeds": { "pairs": { "R2-A4-SPnat-D2-32k - A1-P1r3-D2-16k": { "delta": -0.008684786018473334, "delta_pct": -0.7701554030632566, "ci95_pct": [ -1.2683506028519904, -0.28392151290853485 ], "ci90": [ -0.013318493428309297, -0.004133055845341472 ], "p": 0.0018, "p_plus1": 0.001999800019998, "se_pct": 0.2516827853983878, "ci95_excludes_0": true, "tost_equivalent": false, "mde80_pct": 0.7051107712256217 }, "R2-A10-MinGram-P1r3-D2-32k - A1-P1r3-D2-16k": { "delta": -0.005269543998672743, "delta_pct": -0.4672962320113417, "ci95_pct": [ -0.9419097590776901, 0.042593733706161235 ], "ci90": [ -0.009689398904420075, -0.0005186912539030928 ], "p": 0.0708, "p_plus1": 0.070992900709929, "se_pct": 0.2507693914187916, "ci95_excludes_0": false, "tost_equivalent": false, "mde80_pct": 0.7025518201540716 }, "R2-A10-MinGram-P1r3-D2-48k - A1-P1r3-D2-16k": { "delta": -0.0028933137990660995, "delta_pct": -0.25657526280652554, "ci95_pct": [ -0.5497971900054238, 0.0343370747906838 ], "ci90": [ -0.005623487622485324, -0.00013111641818942617 ], "p": 0.0854, "p_plus1": 0.08559144085591441, "se_pct": 0.14822563760616336, "ci95_excludes_0": false, "tost_equivalent": false, "mde80_pct": 0.415266755262797 }, "R2-A4-SPnat-D2-32k - R2-A10-MinGram-P1r3-D2-32k": { "delta": -0.0034152420198005906, "delta_pct": -0.3042810650033998, "ci95_pct": [ -0.9452945588944334, 0.2739421620049433 ], "ci90": [ -0.00936036830246204, 0.001908147621246279 ], "p": 0.2996, "p_plus1": 0.2997700229977002, "se_pct": 0.307931856086911, "ci95_excludes_0": false, "tost_equivalent": false, "mde80_pct": 0.8626973361991798 }, "R2-A4-SPnat-D2-32k - R2-A10-MinGram-P1r3-D2-48k": { "delta": -0.005791472219407234, "delta_pct": -0.5149012494908062, "ci95_pct": [ -1.019560360215281, -0.03635460358197493 ], "ci90": [ -0.010475266194312948, -0.001285244703859535 ], "p": 0.0356, "p_plus1": 0.0357964203579642, "se_pct": 0.24936744311133924, "ci95_excludes_0": true, "tost_equivalent": false, "mde80_pct": 0.6986241424993546 }, "R2-A10-MinGram-P1r3-D2-48k - R2-A10-MinGram-P1r3-D2-32k": { "delta": 0.0023762301996066437, "delta_pct": 0.21171028338184003, "ci95_pct": [ -0.30725835331838786, 0.7166814161587275 ], "ci90": [ -0.0025137876669922396, 0.007022484614695488 ], "p": 0.4218, "p_plus1": 0.42195780421957807, "se_pct": 0.2593685536234114, "ci95_excludes_0": false, "tost_equivalent": false, "mde80_pct": 0.72664310587469 } }, "p_first_within_top_set": { "R2-A4-SPnat-D2-32k": 0.8448, "R2-A10-MinGram-P1r3-D2-48k": 0.0087, "R2-A10-MinGram-P1r3-D2-32k": 0.1465 }, "what": "clusters resampled (the test cluster draws), each candidate's available seeds averaged, plus relative seed noise N(0, sigma_c^2/n_c); sigma_c = max(dev large 2-seed s.d., dev confirm 5-seed s.d.) from amendment1_stats.json; normal draws from default_rng(12346)" } }