Mathematical Reasoning on GSM8K All Non-Easy
29.7Recovery RateQA+
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| QA+Protocol=opus→opus→haiku2026.02 | 29.7 | 145 | — | — | |
| QAProtocol=haiku→opus→haiku2026.02 | 26.7 | 145 | 195 | 0.31 | |
| Bit-Limited Chain-of-Thought (BL-CoT)Protocol=haiku→haiku→haiku2026.02 | 22.5 | 145 | — | — |