Conversational Question Answering on QuAC-2 2,000
58.05AccuracySINKTRACK
Evaluation results
| Method | Links | ||
|---|---|---|---|
| SINKTRACKModel=Llama3.1-8B-Instruct, Prompting Strategy=SINKTRACK2026.04 | 58.05 | 54.25 | |
| DirectModel=Llama3.1-8B-Instruct, Prompting Strategy=Direct2026.04 | 54.25 | 51.31 | |
| SINKTRACKModel=Qwen2.5-7B-Instruct, Prompting Strategy=SINKTRACK2026.04 | 52.98 | 48.08 | |
| DirectModel=Qwen2.5-7B-Instruct, Prompting Strategy=Direct2026.04 | 52 | 48.06 | |
| CoTModel=Llama3.1-8B-Instruct, Prompting Strategy=CoT2026.04 | 51.75 | 38.21 | |
| SINKTRACKModel=MiniCPM3-4B, Prompting Strategy=SINKTRACK2026.04 | 50.12 | 50.98 | |
| DirectModel=MiniCPM3-4B, Prompting Strategy=Direct2026.04 | 49.97 | 50.94 | |
| CoTModel=MiniCPM3-4B, Prompting Strategy=CoT2026.04 | 41.03 | 43.35 | |
| CoTModel=Qwen2.5-7B-Instruct, Prompting Strategy=CoT2026.04 | 37.3 | 39.42 |