Reasoning on Omni Hard
18.58AccuracyCog-DRIFT
Evaluation results
| Method | Links | |
|---|---|---|
| Cog-DRIFTBase Model=Qwen3-4B-Instruct-25072026.04 | 18.58 | |
| GRPOBase Model=Qwen3-4B-Instruct-25072026.04 | 16.37 | |
| Few-shotBase Model=Qwen3-4B-Instruct-25072026.04 | 16.31 | |
| NuRL (Abstract)Base Model=Qwen3-4B-Instruct-25072026.04 | 16.03 | |
| NuRL (Prefix)Base Model=Qwen3-4B-Instruct-25072026.04 | 15.69 | |
| Zero-shotBase Model=Qwen3-4B-Instruct-25072026.04 | 15.67 | |
| RFTBase Model=Qwen3-4B-Instruct-25072026.04 | 15.13 | |
| NuRL (Prefix)Base Model=Llama3.2-3B-Instruct2026.04 | 7.19 | |
| Cog-DRIFTBase Model=Llama3.2-3B-Instruct2026.04 | 7.04 | |
| NuRL (Abstract)Base Model=Llama3.2-3B-Instruct2026.04 | 6.67 | |
| RFTBase Model=Llama3.2-3B-Instruct2026.04 | 6.52 | |
| Few-shotBase Model=Llama3.2-3B-Instruct2026.04 | 5.74 | |
| Zero-shotBase Model=Llama3.2-3B-Instruct2026.04 | 5.09 | |
| GRPOBase Model=Llama3.2-3B-Instruct2026.04 | 4.32 |