RoboRACE-Net-M / ablation.json
Nabidnur's picture
RACE-Net-M weights, card and code
18ade57 verified
Raw History Blame Contribute Delete
1.66 kB
{
"metric": "accuracy on cells that a peer reported and the receiver has not sensed, 6,111 held-out snapshots (maze seeds 9000+)",
"rows": [
{
"model": "rounds = 1 (no refinement)",
"val_cell_acc": 0.9181,
"val_cell_acc_bma": 0.9014,
"by_liars": {
"0": [
0.9996,
1.0
],
"1": [
0.9492,
0.9398
],
"2": [
0.9003,
0.8725
],
"3": [
0.8566,
0.8269
],
"4": [
0.8295,
0.7985
],
"5": [
0.7606,
0.7293
]
}
},
{
"model": "rounds = 2",
"val_cell_acc": 0.922,
"val_cell_acc_bma": 0.9014,
"by_liars": {
"0": [
0.9997,
1.0
],
"1": [
0.9525,
0.9398
],
"2": [
0.9029,
0.8725
],
"3": [
0.8636,
0.8269
],
"4": [
0.8378,
0.7985
],
"5": [
0.7744,
0.7293
]
}
},
{
"model": "rounds = 3 (released)",
"val_cell_acc": 0.9233,
"val_cell_acc_bma": 0.9014,
"by_liars": {
"0": [
0.9998,
1.0
],
"1": [
0.9542,
0.9398
],
"2": [
0.9035,
0.8725
],
"3": [
0.8673,
0.8269
],
"4": [
0.839,
0.7985
],
"5": [
0.7762,
0.7293
]
}
},
{
"model": "rounds = 3, second seed",
"val_cell_acc": 0.9233,
"val_cell_acc_bma": 0.9014,
"by_liars": {
"0": [
0.9996,
1.0
],
"1": [
0.9539,
0.9398
],
"2": [
0.9042,
0.8725
],
"3": [
0.8665,
0.8269
],
"4": [
0.8401,
0.7985
],
"5": [
0.7764,
0.7293
]
}
}
]
}