Mathematical Reasoning on GSM8K (Score & Speedup)
77.6AccuracyLSP
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| LSPModel=LLaDA-8B, Inference Strategy=LSP2026.03 | 77.6 | 1.51 | |
| FullModel=LLaDA-8B, Inference Strategy=Full2026.03 | 77.1 | — | |
| LSPModel=Dream-7B, Inference Strategy=LSP2026.03 | 75.4 | 1.69 | |
| FullModel=Dream-7B, Inference Strategy=Full2026.03 | 75.3 | — | |
| Llama3.2-1BForgetting Task=none2025.05 | 19.71 | — | |
| LWFForgetting Task=dental2025.05 | 10.4 | — | |
| LWFForgetting Task=mixed2025.05 | 6.95 | — | |
| LWFForgetting Task=qasc2025.05 | 5.38 | — | |
| LWFForgetting Task=sst52025.05 | 2.67 | — | |
| LWFForgetting Task=psychol2025.05 | 1.17 | — |