Conversational Question Answering on QuAC 3,000 3
56.2AccuracySINKTRACK
Evaluation results
| Method | Links | ||
|---|---|---|---|
| SINKTRACKModel=Llama3.1-8B-Instruct, Prompting Strategy=SINKTRACK2026.04 | 56.2 | 53.45 | |
| DirectModel=Llama3.1-8B-Instruct, Prompting Strategy=Direct2026.04 | 53.7 | 51.2 | |
| SINKTRACKModel=Qwen2.5-7B-Instruct, Prompting Strategy=SINKTRACK2026.04 | 52.01 | 47.77 | |
| DirectModel=Qwen2.5-7B-Instruct, Prompting Strategy=Direct2026.04 | 51.26 | 47.64 | |
| CoTModel=Llama3.1-8B-Instruct, Prompting Strategy=CoT2026.04 | 49.13 | 34.47 | |
| SINKTRACKModel=MiniCPM3-4B, Prompting Strategy=SINKTRACK2026.04 | 48.53 | 49.75 | |
| DirectModel=MiniCPM3-4B, Prompting Strategy=Direct2026.04 | 48.13 | 49.53 | |
| CoTModel=MiniCPM3-4B, Prompting Strategy=CoT2026.04 | 39.39 | 41.19 | |
| CoTModel=Qwen2.5-7B-Instruct, Prompting Strategy=CoT2026.04 | 36.34 | 37.92 |