RNAQwen / reports /eval_selfcheck_bert.json
RomanCohort's picture
add reports/eval_selfcheck_bert.json
defe93e verified
Raw History Blame Contribute Delete
5.98 kB
{
"embedder": "C:\\qwen-api\\RNAQwen\\models\\bert-base-uncased",
"A_rouge_parity": {
"available": true,
"cases": [
{
"case": "identical",
"ours": {
"rouge1": 1.0,
"rouge2": 1.0,
"rougeL": 1.0
},
"rouge_score": {
"rouge1": 1.0,
"rouge2": 1.0,
"rougeL": 1.0
},
"max_abs_diff": 0.0,
"ok": true
},
{
"case": "paraphrase",
"ours": {
"rouge1": 0.6666666666666666,
"rouge2": 0.3283582089552239,
"rougeL": 0.46376811594202894
},
"rouge_score": {
"rouge1": 0.6666666666666666,
"rouge2": 0.3283582089552239,
"rougeL": 0.46376811594202894
},
"max_abs_diff": 0.0,
"ok": true
},
{
"case": "partial",
"ours": {
"rouge1": 0.47058823529411764,
"rouge2": 0.326530612244898,
"rougeL": 0.47058823529411764
},
"rouge_score": {
"rouge1": 0.47058823529411764,
"rouge2": 0.326530612244898,
"rougeL": 0.47058823529411764
},
"max_abs_diff": 0.0,
"ok": true
},
{
"case": "unrelated",
"ours": {
"rouge1": 0.10714285714285714,
"rouge2": 0.0,
"rougeL": 0.10714285714285714
},
"rouge_score": {
"rouge1": 0.10714285714285714,
"rouge2": 0.0,
"rougeL": 0.10714285714285714
},
"max_abs_diff": 0.0,
"ok": true
},
{
"case": "empty",
"ours": {
"rouge1": 0.0,
"rouge2": 0.0,
"rougeL": 0.0
},
"rouge_score": {
"rouge1": 0.0,
"rouge2": 0.0,
"rougeL": 0
},
"max_abs_diff": 0.0,
"ok": true
},
{
"case": "punctuation",
"ours": {
"rouge1": 0.34782608695652173,
"rouge2": 0.3181818181818182,
"rougeL": 0.34782608695652173
},
"rouge_score": {
"rouge1": 0.34782608695652173,
"rouge2": 0.3181818181818182,
"rougeL": 0.34782608695652173
},
"max_abs_diff": 0.0,
"ok": true
},
{
"case": "numbers",
"ours": {
"rouge1": 0.8750000000000001,
"rouge2": 0.8571428571428571,
"rougeL": 0.8750000000000001
},
"rouge_score": {
"rouge1": 0.8750000000000001,
"rouge2": 0.8571428571428571,
"rougeL": 0.8750000000000001
},
"max_abs_diff": 0.0,
"ok": true
}
],
"max_abs_diff": 0.0,
"passed": true
},
"B_semantic_invariants": {
"embedder": "C:\\qwen-api\\RNAQwen\\models\\bert-base-uncased",
"cases": [
{
"case": "identical",
"expect": "high",
"precision": 1.0000000173987469,
"recall": 1.0000000173987469,
"f1": 1.0000000173987469,
"embedder": "C:\\qwen-api\\RNAQwen\\models\\bert-base-uncased",
"n": 1,
"n_sent_hyp": 3,
"n_sent_ref": 3
},
{
"case": "paraphrase",
"expect": "high",
"precision": 0.9395574293266001,
"recall": 0.9395574293266001,
"f1": 0.9395574293266001,
"embedder": "C:\\qwen-api\\RNAQwen\\models\\bert-base-uncased",
"n": 1,
"n_sent_hyp": 3,
"n_sent_ref": 3
},
{
"case": "partial",
"expect": "mid",
"precision": 0.9365732525262755,
"recall": 0.8748133303355677,
"f1": 0.9046404272811341,
"embedder": "C:\\qwen-api\\RNAQwen\\models\\bert-base-uncased",
"n": 1,
"n_sent_hyp": 1,
"n_sent_ref": 3
},
{
"case": "unrelated",
"expect": "low",
"precision": 0.8003520631037216,
"recall": 0.8320504559443479,
"f1": 0.8158934959372496,
"embedder": "C:\\qwen-api\\RNAQwen\\models\\bert-base-uncased",
"n": 1,
"n_sent_hyp": 2,
"n_sent_ref": 3
}
],
"passed": true,
"checks": {
"identical_ge_0.999": true,
"paraphrase_gt_partial": true,
"partial_gt_unrelated": true,
"unrelated_lt_0.95": true
}
},
"C_ablation_scale": {
"rouge": {
"rouge1": 0.392,
"rouge2": 0.3833333333333333,
"rougeL": 0.392,
"n": 5,
"beta": 1.0,
"protocol": "rouge_score"
},
"semantic": {
"precision": 0.8186549879282623,
"recall": 0.7539599638912912,
"f1": 0.7843287846903132,
"embedder": "C:\\qwen-api\\RNAQwen\\models\\bert-base-uncased",
"n": 4,
"n_sent_hyp": 6,
"n_sent_ref": 15
},
"per_variant": [
{
"variant": "perfect_100pct",
"rouge1": 1.0,
"rouge2": 1.0,
"rougeL": 1.0,
"semantic_f1": 1.0000000173987469
},
{
"variant": "half_50pct",
"rouge1": 0.4799999999999999,
"rouge2": 0.4583333333333333,
"rougeL": 0.4799999999999999,
"semantic_f1": 0.9369965700408391
},
{
"variant": "one_third",
"rouge1": 0.4799999999999999,
"rouge2": 0.4583333333333333,
"rougeL": 0.4799999999999999,
"semantic_f1": 0.9369965700408391
},
{
"variant": "bare_sequence_like",
"rouge1": 0.0,
"rouge2": 0.0,
"rougeL": 0.0,
"semantic_f1": 0.2633219812808278
},
{
"variant": "empty",
"rouge1": 0.0,
"rouge2": 0.0,
"rougeL": 0.0,
"semantic_f1": 0.0
}
],
"monotonic_rouge1": true
}
}