metric,pair,mean_a,mean_b,diff,ci_lo,ci_hi,p_raw,p_adj,cohen_d,n Mapped rate,Human vs R3,0.4899895193896488,0.7155877536329857,-0.22559823424333678,-0.2676959974866721,-0.18271024365985206,3.091669607272657e-15,1.360334627199969e-13,-0.9391705536002956,142 Mapped rate,Human vs Gemini 3 Pro,0.4899895193896488,0.7097082494969819,-0.21971873010733306,-0.26857579119943703,-0.17067950497644865,2.890550568746666e-13,1.0984092161237332e-11,-0.8601416165838744,142 Mapped rate,Human vs GPT-5.2,0.4899895193896488,0.5618179891418426,-0.07182846975219376,-0.1133164986764651,-0.030054673547188514,0.0004675034791561541,0.009817573062279237,-0.3113678147138991,142 Mapped rate,R3 vs Gemini 3 Pro,0.7155877536329857,0.7097082494969819,0.005879504136003685,-0.02716677598613299,0.03854804982364718,0.4599351604585328,1.0,0.023899644949806856,142 Mapped rate,R3 vs GPT-5.2,0.7155877536329857,0.5618179891418426,0.15376976449114302,0.1258855447925766,0.18110673503204788,6.920158790697418e-17,3.321676219534761e-15,0.6983454525949354,142 Mapped rate,Gemini 3 Pro vs GPT-5.2,0.7097082494969819,0.5618179891418426,0.14789026035513933,0.11209433918751646,0.1837596424116105,1.0122775099758847e-11,3.5429712849155963e-10,0.6247524446201599,142 Consequential comment rate,Human vs R3,0.6748542134664751,0.8817599310208328,-0.20690571755435772,-0.2551748275509831,-0.16073666726169752,4.842408395169645e-13,1.7916911062127686e-11,-1.0388610976289767,138 Consequential comment rate,Human vs Gemini 3 Pro,0.6776946508953231,0.6838199513381995,-0.0061253004428763625,-0.06244350272649708,0.049913078909659035,0.996903647548324,1.0,-0.023387454350337067,137 Consequential comment rate,Human vs GPT-5.2,0.67719339178686,0.8111759705969787,-0.13398257881011855,-0.1831527739595485,-0.08641967606386877,1.5737577449550082e-06,4.8786490093605256e-05,-0.6415430698210449,139 Consequential comment rate,R3 vs Gemini 3 Pro,0.8784490748633924,0.6822619047619046,0.19618717010148753,0.15074970202558574,0.24356392463675364,9.322032131973226e-13,3.3559315675103616e-11,0.9288418601196624,140 Consequential comment rate,R3 vs GPT-5.2,0.8793111381622335,0.8128411543980347,0.0664699837641988,0.03689422154371995,0.09594383761808879,1.0091186073505552e-05,0.00028255321005815544,0.4980351920699379,141 Consequential comment rate,Gemini 3 Pro vs GPT-5.2,0.6822619047619046,0.8115043055008778,-0.12924240073897306,-0.18174005053116205,-0.07885540723773254,1.4571271414234164e-05,0.0003642817853558541,-0.590488723778027,140 Consequential rate (critical only),Human vs R3,0.8347786480451296,0.8991964657542422,-0.06441781770911255,-0.10914881300851381,-0.021849944721013585,0.04837195991430917,0.58046351897171,-0.3561370018059802,136 Consequential rate (critical only),Human vs Gemini 3 Pro,0.8322812523138792,0.9293589743589743,-0.09707772204509522,-0.14443104550000688,-0.050642544550591105,0.00014969807655524226,0.003443055760770572,-0.46497864694598,130 Consequential rate (critical only),Human vs GPT-5.2,0.8359846433148732,0.8788806149018348,-0.042895971586961616,-0.08673352806442312,-0.001932882565735495,0.32812814264751056,1.0,-0.22983345254131454,137 Consequential rate (critical only),R3 vs Gemini 3 Pro,0.8976948967759234,0.9240049751243781,-0.02631007834845458,-0.06167910430129711,0.012962082625785175,0.0030384725215295023,0.05165403286600154,-0.16638817324368502,134 Consequential rate (critical only),R3 vs GPT-5.2,0.900466094628205,0.8813034546006683,0.019162640027536936,-0.005293332285110663,0.04384917685151119,0.15957684526232796,1.0,0.1623975032742464,141 Consequential rate (critical only),Gemini 3 Pro vs GPT-5.2,0.9240049751243781,0.8751028887962256,0.048902086328152586,0.010378051960002889,0.08410262460258706,0.00016086659127317505,0.003539065008009851,0.2970774472377778,134 Specification rate,Human vs R3,0.9900805075161558,1.0,-0.00991949248384411,-0.018328152912599243,-0.0037345283824157063,0.0014688499662158369,0.029376999324316738,-0.3133705706247115,142 Specification rate,Human vs Gemini 3 Pro,0.9900805075161558,1.0,-0.00991949248384411,-0.018185544444043716,-0.003560250391236308,0.0014688499662158369,0.029376999324316738,-0.3133705706247115,142 Specification rate,Human vs GPT-5.2,0.9900805075161558,1.0,-0.00991949248384411,-0.01822885878040079,-0.0036514151961577917,0.0014688499662158369,0.029376999324316738,-0.3133705706247115,142 Specification rate,R3 vs Gemini 3 Pro,1.0,1.0,0.0,0.0,0.0,1.0,1.0,0.0,142 Specification rate,R3 vs GPT-5.2,1.0,1.0,0.0,0.0,0.0,1.0,1.0,0.0,142 Specification rate,Gemini 3 Pro vs GPT-5.2,1.0,1.0,0.0,0.0,0.0,1.0,1.0,0.0,142 Actionability rate,Human vs R3,0.7979423316343255,0.997219357333935,-0.1992770256996095,-0.22959990928291726,-0.1705902873335626,4.9708181860428526e-21,2.6842418204631403e-19,-1.5291853147743861,142 Actionability rate,Human vs Gemini 3 Pro,0.7979423316343255,0.8086519114688128,-0.010709579834487455,-0.04832764615883122,0.02702487760279174,0.5743037164017368,1.0,-0.06250559700778213,142 Actionability rate,Human vs GPT-5.2,0.7979423316343255,0.9553277164373162,-0.15738538480299077,-0.18926706026294327,-0.12780951341980532,9.778770491488865e-18,4.889385245744432e-16,-1.1883666144036371,142 Actionability rate,R3 vs Gemini 3 Pro,0.997219357333935,0.8086519114688128,0.18856744586512203,0.16355198026450077,0.21620995423340955,1.332198963733205e-21,7.593534093279268e-20,1.6773098244244666,142 Actionability rate,R3 vs GPT-5.2,0.997219357333935,0.9553277164373162,0.04189164089661872,0.035648041195239615,0.04822862954451199,6.793597170788987e-18,3.4647345571023835e-16,1.4610726286651452,142 Actionability rate,Gemini 3 Pro vs GPT-5.2,0.8086519114688128,0.9553277164373162,-0.14667580496850333,-0.17483101424674538,-0.12148978528724906,1.8244388494097937e-21,1.0216857556694845e-19,-1.2768158286583038,142 Anchored rate,Human vs R3,0.6534063439929538,0.7794498494026086,-0.1260435054096549,-0.16773975229231228,-0.08565376378879103,1.8019389694474693e-07,5.766204702231902e-06,-0.6632675123062529,142 Anchored rate,Human vs Gemini 3 Pro,0.6534063439929538,0.8628772635814889,-0.2094709195885351,-0.24701466774743686,-0.17215720964021514,1.9025893470185118e-16,8.751910996285155e-15,-1.2540916870621595,142 Anchored rate,Human vs GPT-5.2,0.6534063439929538,0.8680364355223793,-0.21463009152942542,-0.2524865675208859,-0.17738188566926993,1.5198752023199168e-17,7.447388491367592e-16,-1.240338262067341,142 Anchored rate,R3 vs Gemini 3 Pro,0.7794498494026086,0.8628772635814889,-0.0834274141788802,-0.11756035173141026,-0.049507867419720845,1.2024465550979193e-05,0.000312636104325459,-0.5894745466400029,142 Anchored rate,R3 vs GPT-5.2,0.7794498494026086,0.8680364355223793,-0.08858658611977052,-0.12383670963548184,-0.05454786790720698,1.8172008645182867e-06,5.45160259355486e-05,-0.5962396386429285,142 Anchored rate,Gemini 3 Pro vs GPT-5.2,0.8628772635814889,0.8680364355223793,-0.005159171940890312,-0.033081889399801945,0.024368852582912937,0.13485847669629392,1.0,-0.04382152759009635,142 Justification rate,Human vs R3,0.6912264032030893,0.9704596242507957,-0.2792332210477066,-0.3129888350586218,-0.2464125141247753,9.179910654163815e-23,5.691544605581566e-21,-1.8930641091875522,142 Justification rate,Human vs Gemini 3 Pro,0.6912264032030893,0.940710932260228,-0.24948452905713892,-0.2865458736316842,-0.21236276649237923,7.43791228939834e-20,3.8677143904871364e-18,-1.5354733889958776,142 Justification rate,Human vs GPT-5.2,0.6912264032030893,0.8829064334443177,-0.19168003024122882,-0.2300257904136557,-0.1531632933097727,3.35488098283328e-15,1.4425988226183103e-13,-1.1878867399259974,142 Justification rate,R3 vs Gemini 3 Pro,0.9704596242507957,0.940710932260228,0.029748691990567653,0.008240815337078973,0.05258387204270261,0.008910937320541013,0.12475312248757418,0.30819702805526805,142 Justification rate,R3 vs GPT-5.2,0.9704596242507957,0.8829064334443177,0.08755319080647775,0.06515841707690244,0.11043871320150847,2.0846450487923997e-11,7.087793165894159e-10,0.9252127570153522,142 Justification rate,Gemini 3 Pro vs GPT-5.2,0.940710932260228,0.8829064334443177,0.0578044988159101,0.03245989517786741,0.08245484265021093,1.1335509642002617e-05,0.00030605876033407066,0.49571330315889517,142 Claim coverage,Human vs R3,0.5957746478873239,0.8322769953051643,-0.23650234741784035,-0.2814553990610329,-0.19271713615023486,2.7662750259130794e-14,1.078847260106101e-12,-1.000883860923557,142 Claim coverage,Human vs Gemini 3 Pro,0.5957746478873239,0.5362676056338028,0.059507042253521124,0.014084507042253527,0.10587147887323939,0.004822914546655651,0.07234371819983476,0.24864871524572013,142 Claim coverage,Human vs GPT-5.2,0.5957746478873239,0.8585680751173709,-0.2627934272300469,-0.30845363849765256,-0.21795774647887325,3.61307978446047e-15,1.5174935094733977e-13,-1.142853978319419,142 Claim coverage,R3 vs Gemini 3 Pro,0.8322769953051643,0.5362676056338028,0.2960093896713615,0.26291079812206575,0.32981220657276994,3.0570303746937705e-21,1.6813667060815738e-19,1.5075657876171862,142 Claim coverage,R3 vs GPT-5.2,0.8322769953051643,0.8585680751173709,-0.026291079812206575,-0.054812206572769945,0.0010563380281690112,0.09339352446637673,1.0,-0.14226412549857437,142 Claim coverage,Gemini 3 Pro vs GPT-5.2,0.5362676056338028,0.8585680751173709,-0.32230046948356805,-0.3566930751173709,-0.2887323943661972,4.0335585743721844e-22,2.4201351446233107e-20,-1.7083556933588497,142 Comments per paper,Human vs R3,17.676056338028168,18.492957746478872,-0.8169014084507042,-2.647887323943662,1.0774647887323943,0.17550988666053013,1.0,-0.10220164417154688,142 Comments per paper,Human vs Gemini 3 Pro,17.676056338028168,5.401408450704225,12.274647887323944,10.5,14.161971830985916,9.778327244356214e-22,5.671429801726605e-20,1.5733822036075709,142 Comments per paper,Human vs GPT-5.2,17.676056338028168,21.570422535211268,-3.8943661971830985,-5.746478873239437,-1.971830985915493,3.29369945526524e-05,0.0007904878692636577,-0.4816212935836819,142 Comments per paper,R3 vs Gemini 3 Pro,18.492957746478872,5.401408450704225,13.091549295774648,12.661971830985916,13.514084507042254,3.727162961375914e-25,2.459927554508103e-23,7.11917374084505,142 Comments per paper,R3 vs GPT-5.2,18.492957746478872,21.570422535211268,-3.0774647887323945,-3.711443661971831,-2.4364436619718335,2.3974820659155632e-14,9.589928263662252e-13,-1.0947927547857088,142 Comments per paper,Gemini 3 Pro vs GPT-5.2,5.401408450704225,21.570422535211268,-16.169014084507044,-16.661971830985916,-15.67605633802817,4.0536798233514e-25,2.63489188517841e-23,-7.322680584706349,142 Comment length (words),Human vs R3,177.86427258904843,78.20468584895119,99.65958674009724,80.91043183055247,119.71245975941156,5.78044719713081e-22,3.410463846307178e-20,1.1815605475575321,142 Comment length (words),Human vs Gemini 3 Pro,177.86427258904843,63.665274983232734,114.19899760581568,95.70336220402508,134.6147605263404,2.966155581934752e-24,1.8983395724382412e-22,1.3516525015622167,142 Comment length (words),Human vs GPT-5.2,177.86427258904843,68.08690892138931,109.77736366765912,90.91461565325832,129.99423186275456,7.106350276028737e-24,4.477000673898104e-22,1.3003190016492436,142 Comment length (words),R3 vs Gemini 3 Pro,78.20468584895119,63.665274983232734,14.539410865718445,12.873145738301472,16.158446219772564,1.344915263055979e-22,8.203983104641472e-21,1.9719181133006254,142 Comment length (words),R3 vs GPT-5.2,78.20468584895119,68.08690892138931,10.117776927561865,8.714615013213441,11.52155250900332,9.622613525259024e-21,5.099985168387282e-19,1.5361407648214775,142 Comment length (words),Gemini 3 Pro vs GPT-5.2,63.665274983232734,68.08690892138931,-4.421633938156578,-6.187833961937802,-2.6592827227187934,2.2834077858657585e-06,6.6218825790107e-05,-0.5382858242495019,142 Claim overlap fraction with human,Human vs R3,,,,,,,,,0 Claim overlap fraction with human,Human vs Gemini 3 Pro,,,,,,,,,0 Claim overlap fraction with human,Human vs GPT-5.2,,,,,,,,,0 Claim overlap fraction with human,R3 vs Gemini 3 Pro,0.8672661870503598,0.6699040767386092,0.19736211031175063,0.14928057553956833,0.24508393285371705,5.3137418641912065e-11,1.7535348151830982e-09,0.764931339927088,139 Claim overlap fraction with human,R3 vs GPT-5.2,0.8672661870503598,0.933093525179856,-0.0658273381294964,-0.10923561151079138,-0.023860911270983217,0.0032233456399310295,0.05165403286600154,-0.33877458150841894,139 Claim overlap fraction with human,Gemini 3 Pro vs GPT-5.2,0.6699040767386092,0.933093525179856,-0.263189448441247,-0.3110341726618705,-0.21510791366906476,1.3075027970125942e-14,5.360761467751637e-13,-1.1442117574584187,139 Stance-matched overlap fraction with human,Human vs R3,,,,,,,,,0 Stance-matched overlap fraction with human,Human vs Gemini 3 Pro,,,,,,,,,0 Stance-matched overlap fraction with human,Human vs GPT-5.2,,,,,,,,,0 Stance-matched overlap fraction with human,R3 vs Gemini 3 Pro,0.7631894484412469,0.5001199040767386,0.2630695443645084,0.21798561151079135,0.3109142685851318,2.399889919677485e-15,1.0799504638548682e-13,0.8786988043039472,139 Stance-matched overlap fraction with human,R3 vs GPT-5.2,0.7631894484412469,0.8067146282973621,-0.043525179856115113,-0.08357613908872902,-0.0062320143884892504,0.04136911963084565,0.5377985552009934,-0.16132808755686992,139 Stance-matched overlap fraction with human,Gemini 3 Pro vs GPT-5.2,0.5001199040767386,0.8067146282973621,-0.30659472422062356,-0.3576768585131895,-0.25707434052757794,1.2736998359750913e-16,5.9863892290829295e-15,-1.084014012556114,139 Stance agreement (critical),R3 vs Gemini 3 Pro,0.9905913978494624,0.8340053763440859,0.15658602150537648,,,,9.029296647925562e-07,0.7283211119667254,124 Stance agreement (critical),Gemini 3 Pro vs GPT-5.2,0.8207671957671958,0.9626984126984128,-0.14193121693121702,,,,3.7051147227667365e-06,-0.5774768402447156,126 Stance agreement (critical),R3 vs GPT-5.2,0.9910941475826971,0.9679389312977099,0.02315521628498718,,,,0.0590582290905367,0.23265049877607977,131