| { |
| "n_gold": 282, |
| "n_refusal_probes": 100, |
| "method": "programmatic (no LLM judge)", |
| "answers_with_quotes": 0.996, |
| "quotes_traceable_to_passages": 0.875, |
| "quotes_exact_after_normalization": 0.845, |
| "quotes_not_found": 0.108, |
| "answers_all_quotes_verified": 0.794, |
| "citation_rate": 0.972, |
| "citation_matches_passage": 0.825, |
| "correct_refusal_rate": 1.0, |
| "false_refusal_rate": 0.004, |
| "robustness_by_prompt_format": { |
| "native": { |
| "quote_verbatim": 0.888, |
| "citation_matches": 0.678 |
| }, |
| "anythingllm_style": { |
| "quote_verbatim": 0.855, |
| "citation_matches": 0.797 |
| }, |
| "xml_tags": { |
| "quote_verbatim": 0.877, |
| "citation_matches": 0.667 |
| }, |
| "bare_passages_no_metadata": { |
| "quote_verbatim": 0.83, |
| "citation_matches": 0.0 |
| }, |
| "n_per_format": 60 |
| } |
| } |
