Science Question Answering on GPQA (Avg@k, Pass@k)
35.66Avg@kSFT
Evaluation results
| Method | Links | ||
|---|---|---|---|
| SFTBase Model=Qwen2.5-7B-Instruct, RL Stage=DAPO, k=42026.05 | 35.66 | 65.63 | |
| IWBase Model=Qwen2.5-7B-Instruct, RL Stage=DAPO, k=42026.05 | 34.82 | 64.29 | |
| IWBase Model=Llama-3.1-8B-Instruct, RL Stage=DAPO, k=42026.05 | 34.32 | 69.2 | |
| RPSFTBase Model=Qwen2.5-7B-Instruct, RL Stage=DAPO, k=42026.05 | 34.21 | 65.18 | |
| RPSFTBase Model=Llama-3.1-8B-Instruct, RL Stage=DAPO, k=42026.05 | 33.48 | 68.08 | |
| SFTBase Model=Llama-3.1-8B-Instruct, RL Stage=DAPO, k=42026.05 | 32.7 | 65.4 | |
| DFTBase Model=Qwen2.5-7B-Instruct, RL Stage=DAPO, k=42026.05 | 30.47 | 58.04 | |
| IWBase Model=Qwen2.5-3B-Instruct, RL Stage=DAPO, k=42026.05 | 29.41 | 65.4 | |
| RPSFTBase Model=Qwen2.5-3B-Instruct, RL Stage=DAPO, k=42026.05 | 28.13 | 62.05 | |
| SFTBase Model=Qwen2.5-3B-Instruct, RL Stage=DAPO, k=42026.05 | 27.73 | 61.61 | |
| DFTBase Model=Llama-3.1-8B-Instruct, RL Stage=DAPO, k=42026.05 | 26.79 | 57.14 | |
| DFTBase Model=Qwen2.5-3B-Instruct, RL Stage=DAPO, k=42026.05 | 26.45 | 55.13 |