Open-ended event forecasting on OpenEP
19.65SP@1Qwen2.5-3B-Instruct
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| Qwen2.5-3B-InstructAlignment Strategy=+ GRPO2026.04 | 19.65 | 5.49 | 13.33 | 26.67 | 6.92 | 10.58 | 0.76 | |
| Qwen2.5-3B-Instruct + SCATTERBackbone=Qwen2.5-3B-Instruct, Alignment Strategy=+ SCATTER2026.04 | 16.25 | 9.65 | 35.56 | 38.89 | 4.88 | 17.59 | 8.21 | |
| Llama3.2-3B-Instruct + SCATTERBackbone=Llama3.2-3B-Instruct, Alignment Strategy=+ SCATTER2026.04 | 15.69 | 7.36 | 25.56 | 36.67 | 4.67 | 15.3 | 4.9 | |
| GRPOBackbone=Qwen3-7B-Instruct2026.04 | 14.1 | — | — | 17.78 | 4.96 | 6.42 | 0.62 | |
| Llama3.2-3B-InstructAlignment Strategy=+ GRPO2026.04 | 11.11 | 2.85 | 5.56 | 20 | 3.14 | 6.35 | 0.62 | |
| Llama3.2-3B-InstructAlignment Strategy=+ SFT2026.04 | 9.1 | 6.81 | 25.56 | 30 | 2.37 | 11.43 | 6.92 | |
| SCATTERBackbone=Qwen3-7B-Instruct2026.04 | 7.36 | — | — | 27.78 | 2.36 | 10.99 | 8.88 | |
| Llama3.2-3B-InstructAlignment Strategy=Base2026.04 | 7.29 | 18.26 | 33.33 | 18.89 | 2.97 | 9.72 | 5.67 | |
| BaseBackbone=Qwen3-7B-Instruct2026.04 | 7.01 | — | — | 18.89 | 3.34 | 9.62 | 4.09 | |
| GPT-4o-mini2026.04 | 6.87 | 18.82 | 32.22 | 17.78 | 3.06 | 9.28 | 4.08 | |
| Qwen2.5-3B-InstructAlignment Strategy=+ SFT2026.04 | 6.46 | 5 | 20 | 21.11 | 2.02 | 8.38 | 5.52 | |
| Qwen2.5-3B-InstructAlignment Strategy=Base2026.04 | 4.93 | 15 | 32.22 | 17.78 | 2.51 | 10.2 | 5.56 | |
| SFTBackbone=Qwen3-7B-Instruct2026.04 | 4.44 | — | — | 23.33 | 1.36 | 9.83 | 6.82 |