alpha-signal-analysis / data /eval /results_multi_model.json
basilwong's picture
Upload data/eval/results_multi_model.json with huggingface_hub
fa805eb verified
Raw
History Blame Contribute Delete
11.4 kB
{
"Nemotron-7B (SFT + GRPO, Manus Teacher)": {
"n_predictions": 342,
"n_observations": 1540,
"decay_curve": [
{
"horizon": 1,
"ic": 0.15133705937535044,
"p_value": 0.0027323239570524877,
"n": 390
},
{
"horizon": 2,
"ic": 0.11839899617543623,
"p_value": 0.02368422189685372,
"n": 365
},
{
"horizon": 5,
"ic": 0.15673166826769214,
"p_value": 0.005605587004974826,
"n": 311
},
{
"horizon": 10,
"ic": 0.16016765774371008,
"p_value": 0.008372760788665752,
"n": 270
},
{
"horizon": 20,
"ic": 0.1585737792001657,
"p_value": 0.023492936952148647,
"n": 204
}
],
"direction_accuracy_5d": {
"accuracy": 0.5862068965517241,
"n": 261
}
},
"Nemotron-7B (Best-of-4 SFT, Manus Teacher)": {
"n_predictions": 258,
"n_observations": 938,
"decay_curve": [
{
"horizon": 1,
"ic": 0.04071017789473223,
"p_value": 0.5276492216130347,
"n": 243
},
{
"horizon": 2,
"ic": 0.21653544398711053,
"p_value": 0.000999127506884105,
"n": 228
},
{
"horizon": 5,
"ic": 0.13896370904672078,
"p_value": 0.060641815758503324,
"n": 183
},
{
"horizon": 10,
"ic": 0.016262804153897626,
"p_value": 0.8382670785659257,
"n": 160
},
{
"horizon": 20,
"ic": 0.12993920429930514,
"p_value": 0.15032248859824535,
"n": 124
}
],
"direction_accuracy_5d": {
"accuracy": 0.5510204081632653,
"n": 147
}
},
"Nemotron-7B (SFT + DPO, Manus Teacher)": {
"n_predictions": 340,
"n_observations": 1444,
"decay_curve": [
{
"horizon": 1,
"ic": 0.06305758987524085,
"p_value": 0.2333434405036308,
"n": 359
},
{
"horizon": 2,
"ic": 0.05997507359090778,
"p_value": 0.272250882135017,
"n": 337
},
{
"horizon": 5,
"ic": 0.06980694232032451,
"p_value": 0.22957425133553674,
"n": 298
},
{
"horizon": 10,
"ic": 0.07728259750345057,
"p_value": 0.21512915006209438,
"n": 259
},
{
"horizon": 20,
"ic": 0.13642872220850605,
"p_value": 0.05984764039311204,
"n": 191
}
],
"direction_accuracy_5d": {
"accuracy": 0.5656108597285068,
"n": 221
}
},
"Nemotron-7B (SFT + Thinking, Manus Teacher)": {
"n_predictions": 363,
"n_observations": 6688,
"decay_curve": [
{
"horizon": 1,
"ic": -0.023466446987174896,
"p_value": 0.3028239493498633,
"n": 1930
},
{
"horizon": 2,
"ic": 0.025401578358725915,
"p_value": 0.30674052833540394,
"n": 1621
},
{
"horizon": 5,
"ic": -0.010029340325530633,
"p_value": 0.7199848544445703,
"n": 1280
},
{
"horizon": 10,
"ic": 0.0483803712312681,
"p_value": 0.11458139253150973,
"n": 1065
},
{
"horizon": 20,
"ic": 0.08981157990172253,
"p_value": 0.011450491212911315,
"n": 792
}
],
"direction_accuracy_5d": {
"accuracy": 0.518025078369906,
"n": 1276
}
},
"Nemotron-7B (SFT + Bearish, Manus Teacher)": {
"n_predictions": 405,
"n_observations": 7704,
"decay_curve": [
{
"horizon": 1,
"ic": -0.06004980972308451,
"p_value": 0.004522103780931757,
"n": 2234
},
{
"horizon": 2,
"ic": -0.06364464371955256,
"p_value": 0.005680583716863656,
"n": 1887
},
{
"horizon": 5,
"ic": -0.09414103682583536,
"p_value": 0.0002598442555052048,
"n": 1501
},
{
"horizon": 10,
"ic": 0.0410165813555311,
"p_value": 0.15441259574325564,
"n": 1207
},
{
"horizon": 20,
"ic": 0.10274480416130091,
"p_value": 0.0023427591148939392,
"n": 875
}
],
"direction_accuracy_5d": {
"accuracy": 0.48130841121495327,
"n": 1498
}
},
"Nemotron-7B (SFT, Manus Teacher)": {
"n_predictions": 388,
"n_observations": 3265,
"decay_curve": [
{
"horizon": 1,
"ic": 0.03183775000348284,
"p_value": 0.36430795464818033,
"n": 814
},
{
"horizon": 2,
"ic": 0.03171656860131455,
"p_value": 0.38542529940549464,
"n": 751
},
{
"horizon": 5,
"ic": 0.07449665928589075,
"p_value": 0.051820262349632655,
"n": 682
},
{
"horizon": 10,
"ic": 0.12812259772393858,
"p_value": 0.001954608091266253,
"n": 582
},
{
"horizon": 20,
"ic": 0.13577796817862944,
"p_value": 0.004509206112266428,
"n": 436
}
],
"direction_accuracy_5d": {
"accuracy": 0.5293233082706766,
"n": 665
}
},
"Nemotron-7B (SFT, GPT-5.5 Teacher)": {
"n_predictions": 311,
"n_observations": 2236,
"decay_curve": [
{
"horizon": 1,
"ic": -0.012449958147774658,
"p_value": 0.7532489491651047,
"n": 640
},
{
"horizon": 2,
"ic": 0.04937556372078573,
"p_value": 0.25649298900834255,
"n": 530
},
{
"horizon": 5,
"ic": 0.0694607041997462,
"p_value": 0.15631124648713524,
"n": 418
},
{
"horizon": 10,
"ic": 0.09092840152484534,
"p_value": 0.0778496001130096,
"n": 377
},
{
"horizon": 20,
"ic": 0.0997090126631559,
"p_value": 0.10143955322543599,
"n": 271
}
],
"direction_accuracy_5d": {
"accuracy": 0.5456730769230769,
"n": 416
}
},
"Nemotron-7B (SFT + GRPO, GPT-5.5 Teacher)": {
"n_predictions": 321,
"n_observations": 2333,
"decay_curve": [
{
"horizon": 1,
"ic": -0.059233100334789464,
"p_value": 0.12252486835804242,
"n": 681
},
{
"horizon": 2,
"ic": -0.04442037171246653,
"p_value": 0.2846862335515659,
"n": 582
},
{
"horizon": 5,
"ic": 0.0462351420377592,
"p_value": 0.316687090194408,
"n": 471
},
{
"horizon": 10,
"ic": 0.09465903779557103,
"p_value": 0.07364898626683457,
"n": 358
},
{
"horizon": 20,
"ic": 0.0026168814674867595,
"p_value": 0.9677632450224569,
"n": 241
}
],
"direction_accuracy_5d": {
"accuracy": 0.5415778251599147,
"n": 469
}
},
"Manus (Teacher, Direct)": {
"n_predictions": 421,
"n_observations": 4465,
"decay_curve": [
{
"horizon": 1,
"ic": 0.07520290575431239,
"p_value": 0.013135444929827307,
"n": 1087
},
{
"horizon": 2,
"ic": 0.07286858827054352,
"p_value": 0.019459523257903594,
"n": 1028
},
{
"horizon": 5,
"ic": 0.049940940375474994,
"p_value": 0.12340206367824198,
"n": 953
},
{
"horizon": 10,
"ic": -0.08302792201906682,
"p_value": 0.018249333711449746,
"n": 808
},
{
"horizon": 20,
"ic": -0.05719640354421617,
"p_value": 0.16565403855865118,
"n": 589
}
],
"direction_accuracy_5d": {
"accuracy": 0.5453488372093023,
"n": 860
}
},
"GPT-5.5 (Teacher, Direct)": {
"n_predictions": 421,
"n_observations": 786,
"decay_curve": [
{
"horizon": 1,
"ic": 0.22341486850535283,
"p_value": 0.0010269317091460065,
"n": 213
},
{
"horizon": 2,
"ic": 0.16379795800127228,
"p_value": 0.02079235834676979,
"n": 199
},
{
"horizon": 5,
"ic": 0.06197806591113748,
"p_value": 0.4006963654412396,
"n": 186
},
{
"horizon": 10,
"ic": -0.03621800696767055,
"p_value": 0.6801262930580105,
"n": 132
},
{
"horizon": 20,
"ic": 0.052385851150254126,
"p_value": 0.7013924221150697,
"n": 56
}
],
"direction_accuracy_5d": {
"accuracy": 0.725,
"n": 160
}
},
"Nemotron-7B (Base, No Fine-Tuning)": {
"n_predictions": 278,
"n_observations": 2539,
"decay_curve": [
{
"horizon": 1,
"ic": -0.12450125217808267,
"p_value": 0.0005915450558105003,
"n": 758
},
{
"horizon": 2,
"ic": -0.017953408844330117,
"p_value": 0.6557354377722434,
"n": 619
},
{
"horizon": 5,
"ic": -0.10723387476125326,
"p_value": 0.015608197712822108,
"n": 508
},
{
"horizon": 10,
"ic": -0.009266803005239055,
"p_value": 0.8572991005825865,
"n": 379
},
{
"horizon": 20,
"ic": 0.08551424811314301,
"p_value": 0.15729563140518094,
"n": 275
}
],
"direction_accuracy_5d": {
"accuracy": 0.5531496062992126,
"n": 508
}
},
"Nemotron-14B (Base, No Fine-Tuning)": {
"n_predictions": 389,
"n_observations": 4438,
"decay_curve": [
{
"horizon": 1,
"ic": -0.1273644649387756,
"p_value": 4.733436207654943e-06,
"n": 1283
},
{
"horizon": 2,
"ic": -0.10074822966768027,
"p_value": 0.001021218531663946,
"n": 1060
},
{
"horizon": 5,
"ic": -0.07181355454160375,
"p_value": 0.03357021319509315,
"n": 876
},
{
"horizon": 10,
"ic": 0.05476606382334482,
"p_value": 0.14922528546860614,
"n": 695
},
{
"horizon": 20,
"ic": 0.09434363317195438,
"p_value": 0.030828052680672444,
"n": 524
}
],
"direction_accuracy_5d": {
"accuracy": 0.49336550060313633,
"n": 829
}
},
"Nemotron-32B (Base, No Fine-Tuning)": {
"n_predictions": 416,
"n_observations": 5791,
"decay_curve": [
{
"horizon": 1,
"ic": -0.0930898472112231,
"p_value": 9.622608795803237e-05,
"n": 1750
},
{
"horizon": 2,
"ic": -0.012070648554869624,
"p_value": 0.6493731737991248,
"n": 1421
},
{
"horizon": 5,
"ic": 0.02585212403158173,
"p_value": 0.3931880030191568,
"n": 1093
},
{
"horizon": 10,
"ic": 0.03609213312411273,
"p_value": 0.2840260403108237,
"n": 883
},
{
"horizon": 20,
"ic": 0.08442602950404149,
"p_value": 0.03218094587377862,
"n": 644
}
],
"direction_accuracy_5d": {
"accuracy": 0.5199619771863118,
"n": 1052
}
}
}