Theorem Generation on Future Theorem Prediction dataset (test)
0.975Structure ScoreCOMPOSE
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| COMPOSEDecoder=DeepSeek-Math 7B2026.05 | 0.975 | 0.84 | 6.6 | 1.223 | |
| Paper-graph-onlyDecoder=DeepSeek-Math 7B2026.05 | 0.94 | 0.826 | 5.3 | 1.084 | |
| Bag-of-PapersDecoder=DeepSeek-Math 7B2026.05 | 0.9 | 0.822 | 4.6 | 0.962 | |
| Prompt-onlyDecoder=DeepSeek-Math 7B2026.05 | 0.89 | 0.777 | 4.8 | 0.518 | |
| COMPOSEDecoder=Mistral 7B2026.05 | 0.877 | 0.819 | 5 | 0.441 | |
| Bag-of-PapersDecoder=Mistral 7B2026.05 | 0.851 | 0.826 | 4.2 | 0.48 | |
| Text-only (LoRA)Decoder=DeepSeek-Math 7B2026.05 | 0.85 | 0.807 | 14.8 | 2.241 | |
| FutureGenDecoder=External DeepSeek-Math 7B2026.05 | 0.645 | 0.71 | 4.8 | 0.52 | |
| Paper-graph-onlyDecoder=Mistral 7B2026.05 | 0.644 | 0.808 | 5 | 0.625 | |
| ResearchAgentDecoder=GPT-4.02026.05 | 0.64 | 0.814 | 9.6 | 0.968 | |
| GoAIDecoder=External DeepSeek-Math 7B2026.05 | 0.37 | 0.771 | 4.7 | 0.809 | |
| GIANTSDecoder=GIANTS2026.05 | 0.165 | 0.752 | 3.7 | 0.314 | |
| GoAIDecoder=GPT-4.02026.05 | 0.085 | 0.821 | 12.5 | 1.933 | |
| ResearchAgentDecoder=External DeepSeek-Math 7B2026.05 | 0.082 | 0.648 | 5.4 | 0.699 | |
| CoIDecoder=GPT-4.02026.05 | 0.08 | 0.811 | 11.8 | 1.449 |