hindko-tokenizer / eval /test_lm_confirm.csv
junaid008's picture
Add evaluation tables and charts
cc35156 verified
Raw History Blame Contribute Delete
3.4 kB
test_rank_of6,id,role,bundle,n_vocab,params_total,params_non_embedding,test_bytes_per_token,n_seeds,test_mean_bpb,test_seed_bpb,test_seed_sd_rel_pct,test_delta_vs_baseline_pct,test_delta_vs_baseline_ci95_pct,test_p_vs_baseline,dev_mean_bpb,dev_rank_of17,dev_delta_vs_baseline_pct,dev_delta_vs_baseline_ci95_pct,change_dev_to_test_pp,change_dev_to_test_ci95_pp,test_delta_vs_best_of6_pct,test_delta_vs_best_of6_ci95_pct,test_holm_p_vs_best_of6
1,R2-A10-MinGram-P1r3-D2-48k,pre-registered pick (DECISION.md),f54c929ba1ab,49152,11256960,1772928,7.373362,5,1.205428270630087,"[1.2047896494901702, 1.2059034101351314, 1.20633928949385, 1.205412639294452, 1.2046963647368312]",0.05865259252616563,-1.3160397372781005,"[-1.467033141440053, -1.1643594058582694]",0.0,1.2263581740516902,2,-1.5817597957708232,"[-1.8675208666941552, -1.355869455836522]",0.2657200584927226,"[0.018640026346452922, 0.5532296179772263]",,,
2,R2-A10-MinGram-P1r3-D2-32k,top-set member (Amendment 1 fallback),f54c929ba1ab,32768,8112384,1772928,7.274954,5,1.2061960938127751,"[1.2058776331394079, 1.2063087984686824, 1.2093894631968845, 1.2069417847891937, 1.2024627894697058]",0.20645467091462297,-1.253180889269175,"[-1.4797207661189518, -1.03708384604859]",0.0,1.2284434797265598,3,-1.4144089196964234,"[-1.662170488805242, -1.191419602468561]",0.16122803042724843,"[-0.04231045435541847, 0.3860564894084195]",0.06369712751857491,"[-0.15251248513420793, 0.2953023739022093]",0.5822
3,R2-A4-SPnat-D2-32k,released default (AMENDMENT_1.md),f54c929ba1ab,32768,8112192,1772928,7.277326,5,1.2091072993866336,"[1.2129939093356459, 1.2090638021044307, 1.2080341771212149, 1.2062515856466753, 1.2091930227252015]",0.20435856219186863,-1.0148512414859043,"[-1.3337703266599765, -0.719411106580577]",0.0,1.2258361359998111,1,-1.6236546250896033,"[-2.1070539470058685, -1.2360590866135281]",0.608803383603699,"[0.172267052994226, 1.1171782340898435]",0.30520511640428255,"[-0.019494874894211908, 0.6262727704237294]",0.1296
4,R2-A1-P1r3-D2-48k,equal-vocabulary BPE control for the 48k pick (report-only),94175d26497f,49152,11257152,1772928,7.352626,5,1.211112897205615,"[1.2092966317143177, 1.2119648668982443, 1.2129522019294052, 1.2110089578596441, 1.210341827626464]",0.11686507492736961,-0.850660355728833,"[-1.0468293493009588, -0.6530838548134663]",0.0,1.2337413764600247,5,-0.9892397608619384,"[-1.2117041492054477, -0.7787421586001484]",0.1385794051331054,"[-0.10002961497460199, 0.38457145452512503]",0.4715856359131806,"[0.31370175058109206, 0.6434340570746931]",0.0
5,A1-P1r3-D2-32k,round-1 confirmatory winner (report-only),94175d26497f,32768,8112576,1772928,7.246036,5,1.2145897732763742,"[1.2157356635342742, 1.2152889816577066, 1.2150796722014732, 1.213604199219369, 1.2132403497690487]",0.09051181811235785,-0.5660213536703774,"[-0.7539943802074569, -0.38067108247804304]",0.0,1.2359686388208577,6,-0.8104965138527125,"[-1.0007432256783066, -0.6574930429130209]",0.24447516018233506,"[0.025092332962363524, 0.48592810710440854]",0.760020556137981,"[0.5867749515939401, 0.9398737865247105]",0.0
6,A1-P1r3-D2-16k,baseline (standard recipe),f54c929ba1ab,16384,4969536,1772928,6.948493,5,1.2215037453107163,"[1.2209580951297607, 1.2201410927166643, 1.2220834417794444, 1.2208247446814235, 1.2235113522462882]",0.10816063260614522,,,,1.2460679763292413,9,,,,,1.3335903157660747,"[1.1780764498105485, 1.4888754375437618]",0.0