Effectiveness scoring on AMI Corpus (All Meetings)
0.6445Spearman's ρQwen3-32B
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Qwen3-32BReasoning mode=non-reasoning, Input configuration=ground truth inputs, Context window size=12026.04 | 0.6445 | 0.4803 | |
| GPT-4oInput configuration=ground truth inputs, Context window size=12026.04 | 0.6341 | 0.4756 | |
| DeepSeek-R1-70BInput configuration=ground truth inputs, Context window size=12026.04 | 0.6132 | 0.4663 | |
| Qwen3-32BReasoning mode=reasoning, Input configuration=ground truth inputs, Context window size=12026.04 | 0.6113 | 0.4578 | |
| Llama3.3-70B-InstructInput configuration=ground truth inputs, Context window size=12026.04 | 0.6072 | 0.4854 | |
| Gemini-2.5-FlashInput configuration=ground truth inputs, Context window size=12026.04 | 0.5624 | 0.4122 |