Property-based retrieval on Reasoning Traces (test)
40.9MAPSAE
Evaluation results
| Method | Links | ||
|---|---|---|---|
| SAELLM reranking=false, n_phrases=18, T=0.22025.12 | 40.9 | 62.7 | |
| SAELLM reranking=true, n_phrases=18, T=0.22025.12 | 40.8 | 61.3 | |
| CombinedLLM reranking=false, n_phrases=18, T=0.22025.12 | 39.6 | 63 | |
| CombinedLLM reranking=true, n_phrases=18, T=0.22025.12 | 39.5 | 63.3 | |
| OpenAI+LLMLLM reranking=true, n_phrases=18, T=0.22025.12 | 33.7 | 56.3 | |
| OpenAI+LLMLLM reranking=false, n_phrases=18, T=0.22025.12 | 33.3 | 52.7 |