Argument Relation Classification on EP debates
74AccuracyClaude Sonnet 4.5
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Claude Sonnet 4.5Evaluation protocol=zero-shot2026.04 | 74 | 78 | |
| Claude Haiku 4.5Evaluation protocol=few-shot2026.04 | 69 | 75 | |
| Claude Haiku 4.5Evaluation protocol=zero-shot2026.04 | 68 | 73 | |
| Claude Haiku 3.5Evaluation protocol=zero-shot2026.04 | 65 | 71 | |
| Qwen3 30BEvaluation protocol=zero-shot2026.04 | 64 | 72 | |
| Claude Haiku 3.5Evaluation protocol=few-shot2026.04 | 55 | 62 | |
| Qwen3 30BEvaluation protocol=few-shot2026.04 | 45 | 54 | |
| Llama 3.1 8BEvaluation protocol=fine-tuned2026.04 | 21 | 27 | |
| Claude Sonnet 4.5Evaluation protocol=zero-shot2026.04 | 0.74 | 0.78 | |
| Claude Haiku 4.5Evaluation protocol=zero-shot2026.04 | 0.68 | 0.73 | |
| Qwen3Evaluation protocol=zero-shot, Model scale=30B2026.04 | 0.64 | 0.72 | |
| Llama 3.1Evaluation protocol=fine-tuned, Model scale=8B2026.04 | 0.21 | 0.27 |