Mathematical Reasoning on GSM8K (test) (Latency Breakdown)
15.9Query Prefill Latency (ms)Explicit CoT
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Explicit CoTBackbone=LLAMA-3.2-3B-INSTRUCT2026.06 | 15.9 | 338.8 | 29.5 | 384.2 | |
| CODIBackbone=LLAMA-3.2-3B-INSTRUCT2026.06 | 16.5 | 88.2 | 30.4 | 135.1 | |
| LOTUSBackbone=LLAMA-3.2-3B-INSTRUCT2026.06 | 16.7 | 133 | 31.5 | 181.2 | |
| SIM-CoTBackbone=LLAMA-3.2-3B-INSTRUCT2026.06 | 29.5 | 162.7 | 59.2 | 251.4 |