Files
claim-extractor-detective-q…/benchmarks/base_benchmark.json

16 lines
421 B
JSON
Raw Normal View History

{
"model": "unsloth/Qwen2.5-3B-Instruct-bnb-4bit (BASE, untrained)",
"metrics": {
"n": 15,
"validity": 1.0,
"avg_claims": 2.067,
"avg_verification_questions": 3.71,
"pct_claims_with_3plus_questions": 1,
"contradiction_recall": 1.0,
"date_extraction_rate": 1.0,
"evidence_verbatim_rate": 1.0,
"avg_claim_words": 5.548,
"sarcasm_handling": 0.75,
"avg_latency_s": 30.214
}
}