warlockee's picture
Upload folder using huggingface_hub
bd3ae0e verified
{
"total_experiments": 1337,
"backbone": {
"VJepa2": {
"count": 801,
"best_ap": 0.9245057778280362,
"mean_ap_top50": 0.895848587214344,
"ci_95_lo": 0.8931908088651325,
"ci_95_hi": 0.8988026291685374,
"median_ap": 0.8139507495399437,
"std_ap": 0.163757339706533,
"mean_ap_all": 0.7612305749966171
},
"DINOv3-L": {
"count": 75,
"best_ap": 0.8235769575519882,
"mean_ap_top50": 0.7560700395011171,
"ci_95_lo": 0.7451405416678706,
"ci_95_hi": 0.766703547343837,
"median_ap": 0.7277818601941584,
"std_ap": 0.21752302704391038,
"mean_ap_all": 0.6318233114990187
},
"Multi-Backbone": {
"count": 9,
"best_ap": 0.7751574573174227,
"mean_ap_top50": 0.44404063462364335,
"ci_95_lo": 0.37414945622298357,
"ci_95_hi": 0.5456917995264369,
"median_ap": 0.37899548133051186,
"std_ap": 0.13206704391541094,
"mean_ap_all": 0.4440406346236434
},
"SigLIP2": {
"count": 51,
"best_ap": 0.7392316895227585,
"mean_ap_top50": 0.3981748695470341,
"ci_95_lo": 0.3401127683819367,
"ci_95_hi": 0.451701948138425,
"median_ap": 0.4774945177309499,
"std_ap": 0.2075058400084436,
"mean_ap_all": 0.39036751916375895
},
"DINOv2": {
"count": 29,
"best_ap": 0.7103549261062475,
"mean_ap_top50": 0.5037299968602356,
"ci_95_lo": 0.46227656211131746,
"ci_95_hi": 0.543598789782232,
"median_ap": 0.4862485765539094,
"std_ap": 0.11189646912003579,
"mean_ap_all": 0.5037299968602358
},
"DINOv3-B": {
"count": 370,
"best_ap": 0.55548206295949,
"mean_ap_top50": 0.2930145939162871,
"ci_95_lo": 0.26554338049159265,
"ci_95_hi": 0.32262923156603673,
"median_ap": 0.08160469972835721,
"std_ap": 0.08535161457981126,
"mean_ap_all": 0.10942303649347601
}
},
"encoder": {
"Zipformer": {
"count": 907,
"best_ap": 0.9245057778280362,
"mean_ap_top50": 0.8929179131680226,
"ci_95_lo": 0.8900310735015892,
"ci_95_hi": 0.8961327947345192,
"median_ap": 0.759955961193099,
"std_ap": 0.32479062831314237,
"mean_ap_all": 0.5619710227983473
},
"Retention": {
"count": 209,
"best_ap": 0.9132446147493699,
"mean_ap_top50": 0.8345888794619971,
"ci_95_lo": 0.8270611428745678,
"ci_95_hi": 0.8428017512884284,
"median_ap": 0.6242232137315284,
"std_ap": 0.3134103524865099,
"mean_ap_all": 0.49491134437383116
},
"Hybrid R-M": {
"count": 45,
"best_ap": 0.9054196697190158,
"mean_ap_top50": 0.6585987816596492,
"ci_95_lo": 0.5851145813861701,
"ci_95_hi": 0.7284871023720563,
"median_ap": 0.7899659247667653,
"std_ap": 0.2491655153104344,
"mean_ap_all": 0.6585987816596494
},
"BiMamba": {
"count": 176,
"best_ap": 0.9016478432286867,
"mean_ap_top50": 0.8447304495061743,
"ci_95_lo": 0.8384383025952225,
"ci_95_hi": 0.8513411823686758,
"median_ap": 0.7493739596505815,
"std_ap": 0.33428322918572545,
"mean_ap_all": 0.5378984388035183
}
},
"loss": {
"Focal (g>=2)": {
"count": 755,
"best_ap": 0.9245057778280362,
"mean_ap_top50": 0.8953384816206035,
"ci_95_lo": 0.8925849777610081,
"ci_95_hi": 0.8983818808582945,
"median_ap": 0.782226760456155,
"std_ap": 0.3257799565308995,
"mean_ap_all": 0.5848799399241609
},
"Focal (g<2)": {
"count": 351,
"best_ap": 0.897879901068386,
"mean_ap_top50": 0.8486286434090803,
"ci_95_lo": 0.8437910091844113,
"ci_95_hi": 0.8539084554234181,
"median_ap": 0.6735741244213165,
"std_ap": 0.32152332436205083,
"mean_ap_all": 0.5070450348226136
},
"BCE": {
"count": 217,
"best_ap": 0.8763074803085055,
"mean_ap_top50": 0.822679259993163,
"ci_95_lo": 0.8169120522883302,
"ci_95_hi": 0.8289150095113637,
"median_ap": 0.6706249786094945,
"std_ap": 0.31550664970236075,
"mean_ap_all": 0.5129708802006305
}
},
"pooling": {
"Attention": {
"count": 1337,
"best_ap": 0.9245057778280362,
"mean_ap_top50": 0.895848587214344,
"ci_95_lo": 0.8931908088651325,
"ci_95_hi": 0.8988026291685374,
"median_ap": 0.7463310750227221,
"std_ap": 0.3235763129430968,
"mean_ap_all": 0.5515716223308416
}
},
"pairwise_tests": {
"backbone": {
"VJepa2 vs DINOv3-B": {
"U_statistic": 293808.0,
"p_value": 2.4175896781190874e-161,
"n_a": 801,
"n_b": 370
},
"VJepa2 vs DINOv3-L": {
"U_statistic": 47840.0,
"p_value": 1.9629452854086958e-17,
"n_a": 801,
"n_b": 75
},
"DINOv3-B vs DINOv3-L": {
"U_statistic": 746.0,
"p_value": 3.1583688286042994e-38,
"n_a": 370,
"n_b": 75
}
},
"encoder": {
"Zipformer vs Retention": {
"U_statistic": 110501.0,
"p_value": 0.00018250872229140202,
"n_a": 907,
"n_b": 209
},
"Zipformer vs BiMamba": {
"U_statistic": 84477.5,
"p_value": 0.21966270517029907,
"n_a": 907,
"n_b": 176
},
"Zipformer vs Hybrid R-M": {
"U_statistic": 16783.0,
"p_value": 0.04412500835643444,
"n_a": 907,
"n_b": 45
},
"Retention vs BiMamba": {
"U_statistic": 16195.0,
"p_value": 0.04345703926564611,
"n_a": 209,
"n_b": 176
}
},
"loss": {
"Focal (g>=2) vs BCE": {
"U_statistic": 102227.5,
"p_value": 2.5147856868233184e-08,
"n_a": 755,
"n_b": 217
},
"Focal (g>=2) vs Focal (g<2)": {
"U_statistic": 162071.0,
"p_value": 2.2293780660306323e-09,
"n_a": 755,
"n_b": 351
}
},
"pooling": {}
}
}