Theme Detection on Colorado
84.8Macro F1 ScoreContext-based strategy
Evaluation Results
| Method | Links | |
|---|---|---|
| Context-based strategyLLM=Mistral2026.03 | 84.8 | |
| Self-DebiasLLM=Gemma2026.03 | 84.2 | |
| Context-based strategyLLM=Gemma2026.03 | 84.2 | |
| Zero-CoTLLM=Mistral2026.03 | 83.6 | |
| Context-based strategyLLM=Llama32026.03 | 83.2 | |
| Self-DebiasLLM=Mistral2026.03 | 83 | |
| Zero-CoTLLM=Gemma2026.03 | 82.6 | |
| Self-DebiasLLM=Llama32026.03 | 79.5 | |
| Zero-CoTLLM=Llama32026.03 | 79.4 |