Relation Extraction on Re-DocRED full (test)
74.3Micro-F1 (Positive Class)Best SLM
Evaluation results
| Method | Links | |
|---|---|---|
| Best SLMPrompting=schema-enumerated, Selection=strongest tuned SLM per benchmark2026.06 | 74.3 | |
| RoB-baseBackbone=RoBERTa, Parameters=125M, Inference mode=fine-tuned per benchmark2026.06 | 71.3 | |
| RoB-largeBackbone=RoBERTa, Parameters=355M, Inference mode=fine-tuned per benchmark2026.06 | 68.2 | |
| GPT-5.4Inference mode=zero-shot2026.06 | 55.9 | |
| ClaudeInference mode=zero-shot2026.06 | 52.3 |