qev / release-comparison /comparison.json
ken-jo's picture
Release QEV 0.1.1: Qwen3.5-2B typed decisions
f996835 verified
Raw History Blame Contribute Delete
3.89 kB
{
"passed": true,
"models": {
"qev": {
"evaluation_sha256": "18e600d2762c06c4ecc7aa07a08055a1a67ca121ec53776e081faed74526ca82",
"predictions_sha256": "2b404c1dccf9459438ebb9303929b1368678307a5d72fdbd2fb82b44ebbba100"
},
"laya-base": {
"evaluation_sha256": "5c69751621bd4a5f00262943ddbc087b3cc68fd00d957ab709547c7a17cf9a95",
"predictions_sha256": "c769919177ae6fb1248e21d0dde16c6b67c473bc068c37631a95f3dceafb9d70"
},
"laya-specialist": {
"evaluation_sha256": "3ebfc3b3a38fb1d30616feea28376e60ebcc4b0ae935cf3a84ba4a95e21ff79c",
"predictions_sha256": "885cb7317dbe422bcb2b05ff2d93f50273ab3d373797d59d556cadf851964cfe"
}
},
"paired_accuracy": {
"laya-base": {
"typed_decisions": {
"qev_minus_reference_accuracy": 0.4095,
"paired_group_bootstrap_95_ci": [
0.3745,
0.4435
],
"groups": 400,
"questions": 2000,
"bootstrap_replicates": 10000,
"seed": 20261001,
"descriptive_not_multiple_comparison_adjusted": true,
"qev_only_correct": 993,
"reference_only_correct": 174
},
"ag_news": {
"qev_minus_reference_accuracy": -0.125,
"paired_group_bootstrap_95_ci": [
-0.1625,
-0.0875
],
"groups": 400,
"questions": 400,
"bootstrap_replicates": 10000,
"seed": 20261001,
"descriptive_not_multiple_comparison_adjusted": true,
"qev_only_correct": 10,
"reference_only_correct": 60
},
"emotion": {
"qev_minus_reference_accuracy": -0.085,
"paired_group_bootstrap_95_ci": [
-0.1325,
-0.035
],
"groups": 400,
"questions": 400,
"bootstrap_replicates": 10000,
"seed": 20261001,
"descriptive_not_multiple_comparison_adjusted": true,
"qev_only_correct": 34,
"reference_only_correct": 68
}
},
"laya-specialist": {
"typed_decisions": {
"qev_minus_reference_accuracy": 0.0005,
"paired_group_bootstrap_95_ci": [
-0.0185,
0.0195
],
"groups": 400,
"questions": 2000,
"bootstrap_replicates": 10000,
"seed": 20261001,
"descriptive_not_multiple_comparison_adjusted": true,
"qev_only_correct": 166,
"reference_only_correct": 165
},
"ag_news": {
"qev_minus_reference_accuracy": -0.1275,
"paired_group_bootstrap_95_ci": [
-0.1675,
-0.09
],
"groups": 400,
"questions": 400,
"bootstrap_replicates": 10000,
"seed": 20261001,
"descriptive_not_multiple_comparison_adjusted": true,
"qev_only_correct": 10,
"reference_only_correct": 61
},
"emotion": {
"qev_minus_reference_accuracy": -0.0975,
"paired_group_bootstrap_95_ci": [
-0.1475,
-0.0475
],
"groups": 400,
"questions": 400,
"bootstrap_replicates": 10000,
"seed": 20261001,
"descriptive_not_multiple_comparison_adjusted": true,
"qev_only_correct": 33,
"reference_only_correct": 72
}
}
},
"requests_sha256": "585850d575d9dbdf943683936dda7e07b9e5525b60875906926e9915c557d1d6",
"input_protocol_sha256": "98b677278157506b94c52c5d449d1f6b805279914e92542781952b42477de4aa",
"aggregator_sha256": "e5802fd38ccdee3d85718faa1c5ac664f0173ed3e84115342445a6712f09d7d2",
"same_cases_questions_labels_and_metrics": true,
"all_input_truncation_counts_zero": true,
"new_training_or_test_temperature_fitting": false,
"scope": "Paired release regression and two task-held-out English suites, not general parity"
}