General Question Answering on TriviaQA (EM, FBERT)
66.3EMCoT
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| CoTSLM=Qwen3-235B2026.04 | 66.3 | 85.95 | |
| Dynamic Collaboration FrameworkSLM=Qwen3-8B2026.04 | 64.64 | 85.32 | |
| Dynamic Collaboration FrameworkSLM=Llama-3.1-8B2026.04 | 62.25 | 84.06 | |
| Dynamic Collaboration FrameworkSLM=Qwen3-4B2026.04 | 61.82 | 83.95 | |
| Direct InferSLM=Qwen3-235B2026.04 | 61.72 | 83.72 | |
| Dynamic Collaboration FrameworkSLM=Llama-3.2-3B2026.04 | 57.5 | 82.01 | |
| CoTSLM=Qwen3-8B2026.04 | 47.83 | 76.19 | |
| CoTSLM=Llama-3.1-8B2026.04 | 46 | 73.44 | |
| CoTSLM=Qwen3-4B2026.04 | 37.57 | 71.21 | |
| CoTSLM=Llama-3.2-3B2026.04 | 34.25 | 67.23 |