Mathematical Reasoning on SVAMP (Pass@1, #L)
60.78Pass@1GRPO
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| GRPOBase Model=LLaMA-3.2-1B-Instruct, Reasoning Mode=Explicit Reasoning2026.04 | 60.78 | 69.4 | |
| SFTBase Model=LLaMA-3.2-1B-Instruct, Reasoning Mode=Explicit Reasoning2026.04 | 57.38 | 44.1 | |
| Latent-GRPOBase Model=LLaMA-3.2-1B-Instruct, Reasoning Mode=Latent Reasoning (No Sampling)2026.04 | 56.5 | 15.8 | |
| Latent-SFTBase Model=LLaMA-3.2-1B-Instruct, Reasoning Mode=Latent Reasoning (No Sampling)2026.04 | 53.22 | 15.3 | |
| Soft-GRPOBase Model=LLaMA-3.2-1B-Instruct, Reasoning Mode=Latent Reasoning (No Sampling)2026.04 | 53.19 | 15.3 |