End-to-end Bridging Resolution on ISNotes
40.2Recognition PrecisionKobayashi et al. (2023)
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Kobayashi et al. (2023)Evaluation Setting=End-to-end2026.05 | 40.2 | 39.5 | 39.9 | 26.4 | 25.9 | 26.2 | |
| LLMBridgeBackend Model=Gemini-3.1-pro-preview, Evaluation Setting=End-to-end2026.05 | 38.5 | 55.7 | 45.6 | 25.2 | 36.5 | 29.8 | |
| LLMBridgeBackend Model=Qwen2.5-7B-Instruct-FT, Evaluation Setting=End-to-end2026.05 | 27.8 | 30.9 | 29.3 | 9.2 | 10.3 | 9.7 | |
| LLMBridgeBackend Model=Llama-3.3-70B-Instruct, Evaluation Setting=End-to-end2026.05 | 19 | 46.4 | 26.9 | 7.7 | 18.7 | 10.9 | |
| LLMBridgeBackend Model=Qwen2.5-7B-Instruct, Evaluation Setting=End-to-end2026.05 | 18.3 | 20.7 | 19.4 | 3.1 | 3.5 | 3.2 |