Multi-Hop Question Answering on MuSiQuE (EM, FBERT)
23.62EMCoT
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| CoTSLM=Qwen3-235B2026.04 | 23.62 | 66.23 | |
| Dynamic Collaboration FrameworkSLM=Qwen3-8B2026.04 | 23.31 | 67.34 | |
| Dynamic Collaboration FrameworkSLM=Qwen3-4B2026.04 | 18.61 | 65 | |
| Direct InferSLM=Qwen3-235B2026.04 | 18.15 | 62.46 | |
| Dynamic Collaboration FrameworkSLM=Llama-3.1-8B2026.04 | 17.79 | 63.7 | |
| Dynamic Collaboration FrameworkSLM=Llama-3.2-3B2026.04 | 14.47 | 60.61 | |
| CoTSLM=Qwen3-8B2026.04 | 12.93 | 58.38 | |
| CoTSLM=Qwen3-4B2026.04 | 10.12 | 57.36 | |
| CoTSLM=Llama-3.1-8B2026.04 | 9.41 | 50.88 | |
| CoTSLM=Llama-3.2-3B2026.04 | 7.11 | 50.02 |