Download summaries/Hermes-3-Llama-3.1-8B_structured_extraction_report.txt from LatamBoard/leaderboard-results: direct link, hf CLI and curl.
- Browser
- Download file 4.38 kB
-
https://huggingface.co/datasets/LatamBoard/leaderboard-results/resolve/main/summaries/Hermes-3-Llama-3.1-8B_structured_extraction_report.txt
- Command line
-
hf download hf://datasets/LatamBoard/leaderboard-results/summaries/Hermes-3-Llama-3.1-8B_structured_extraction_report.txt
-
curl -L -o Hermes-3-Llama-3.1-8B_structured_extraction_report.txt https://huggingface.co/datasets/LatamBoard/leaderboard-results/resolve/main/summaries/Hermes-3-Llama-3.1-8B_structured_extraction_report.txt
4.38 kB
| ┌──────────────────────────────────────────────────────────────────────┐ | |
| │ Model: NousResearch/Hermes-3-Llama-3.1-8B │ | |
| │ Dataset: paraloq/json_data_extraction (476 samples) │ | |
| │ Date: 2025-10-11 01:15:44 │ | |
| └──────────────────────────────────────────────────────────────────────┘ | |
| OVERALL SCORE | |
| ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ | |
| Extraction Quality Score (EQS): 0.849 ✅ GOOD | |
| 95% CI: [0.832, 0.867] | |
| PRIMARY METRICS | |
| ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ | |
| ✅ Schema Validity: 99.4% (473/476) Grade: A+ | |
| ⚠️ Field F1 (Partial): 0.771 Grade: B | |
| ❌ Exact Match: 27.5% (131/476) Grade: D | |
| ✅ Hallucination Rate: 6.6% Grade: A | |
| DETAILED ACCURACY | |
| ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ | |
| Field F1 (Strict): 0.758 | |
| Field Precision: 0.780 | |
| Field Recall: 0.771 | |
| Type Accuracy: 90.4% | |
| MATCH DISTRIBUTION | |
| ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ | |
| Exact 66.9% ██████████████████████████ | |
| Partial 2.7% █ | |
| Incorrect 13.3% █████ | |
| Missed 9.2% ███ | |
| Spurious 7.8% ███ | |
| COMPOSITE SCORE STATISTICS | |
| ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ | |
| Mean: 0.892 | |
| Median: 1.000 | |
| StdDev: 0.275 | |
| Range: [0.003, 1.000] | |
| SAMPLE-LEVEL CONSISTENCY | |
| ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ | |
| EQS: μ=0.849, σ=0.197, range=[0.200, 1.000] | |
| F1: μ=0.771, σ=0.282, range=[0.000, 1.000] | |
| Consistency: Poor (highly variable) | |
| SCHEMA COMPLEXITY ANALYSIS | |
| ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ | |
| Distribution by Complexity: | |
| Simple (443): EQS=0.846, F1=0.767, Halluc= 7.0% | |
| Medium ( 29): EQS=0.898, F1=0.837, Halluc= 2.1% | |
| Complex ( 1): EQS=0.892, F1=0.820, Halluc= 0.0% | |
| Correlations with Performance: | |
| complexity_score_vs_eqs r=+0.183 (negligible positive) | |
| total_fields_vs_f1 r=+0.171 (negligible positive) | |
| max_depth_vs_f1 r=+0.078 (negligible positive) | |
| PRODUCTION READINESS | |
| ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ | |
| Quality Tier: Tier 2 (Good) | |
| Production Ready: ✅ YES | |
| Recommendation: DEPLOY WITH CONFIDENCE | |
| Model demonstrates strong extraction capability across | |
| the benchmark. Suitable for production use. | |
| PERFORMANCE | |
| ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ | |
| Throughput: 0.53 samples/second | |
| Total Duration: 906.54 seconds | |
| Avg per Sample: 1.90 seconds | |