Preference-aligned Retrieval-Augmented Generation on PrefEval
77.96AccuracyEPIC
Evaluation Results
| Method | Links | |
|---|---|---|
| EPICLLM Backend=gpt-oss-20b2026.05 | 77.96 | |
| EPICLLM Backend=Llama-3.1-8B-Instruct2026.05 | 65.61 | |
| EPICLLM Backend=Qwen3-4B-Instruct-25072026.05 | 61.93 | |
| RAPTORLLM Backend=Qwen3-4B-Instruct-2507, Method Category=Indexing-enhanced RAG2026.05 | 38.07 | |
| Pref-QRLLM Backend=Llama-3.1-8B-Instruct, Method Category=Preference-conditioned RAG2026.05 | 34.21 | |
| ContrieverLLM Backend=Qwen3-4B-Instruct-2507, Method Category=Standard RAG2026.05 | 33.51 | |
| HippoRAG 2LLM Backend=Qwen3-4B-Instruct-2507, Method Category=Indexing-enhanced RAG2026.05 | 33.51 | |
| NV-Embed-v2LLM Backend=Qwen3-4B-Instruct-2507, Method Category=Standard RAG2026.05 | 32.98 | |
| HippoRAG 2LLM Backend=Llama-3.1-8B-Instruct, Method Category=Indexing-enhanced RAG2026.05 | 32.98 | |
| PBRLLM Backend=Qwen3-4B-Instruct-2507, Method Category=Preference-conditioned RAG2026.05 | 32.98 | |
| Pref-QRLLM Backend=Qwen3-4B-Instruct-2507, Method Category=Preference-conditioned RAG2026.05 | 32.63 | |
| RAPTORLLM Backend=Llama-3.1-8B-Instruct, Method Category=Indexing-enhanced RAG2026.05 | 31.05 | |
| BM25LLM Backend=Qwen3-4B-Instruct-2507, Method Category=Standard RAG2026.05 | 30.92 | |
| NV-Embed-v2LLM Backend=Llama-3.1-8B-Instruct, Method Category=Standard RAG2026.05 | 30.88 | |
| PBRLLM Backend=Llama-3.1-8B-Instruct, Method Category=Preference-conditioned RAG2026.05 | 30.7 | |
| ContrieverLLM Backend=gpt-oss-20b, Method Category=Standard RAG2026.05 | 29.84 | |
| Pref-QRLLM Backend=gpt-oss-20b, Method Category=Preference-conditioned RAG2026.05 | 29.47 | |
| NV-Embed-v2LLM Backend=gpt-oss-20b, Method Category=Standard RAG2026.05 | 29.28 | |
| RAPTORLLM Backend=gpt-oss-20b, Method Category=Indexing-enhanced RAG2026.05 | 28.07 | |
| BM25LLM Backend=Llama-3.1-8B-Instruct, Method Category=Standard RAG2026.05 | 27.97 | |
| ContrieverLLM Backend=Llama-3.1-8B-Instruct, Method Category=Standard RAG2026.05 | 27.89 | |
| PBRLLM Backend=gpt-oss-20b, Method Category=Preference-conditioned RAG2026.05 | 27.54 | |
| HippoRAG 2LLM Backend=gpt-oss-20b, Method Category=Indexing-enhanced RAG2026.05 | 27.19 | |
| BM25LLM Backend=gpt-oss-20b, Method Category=Standard RAG2026.05 | 26.98 | |
| HippoRAGLLM Backend=gpt-oss-20b, Method Category=Indexing-enhanced RAG2026.05 | 26.37 | |
| HippoRAGLLM Backend=Llama-3.1-8B-Instruct, Method Category=Indexing-enhanced RAG2026.05 | 21.4 | |
| HippoRAGLLM Backend=Qwen3-4B-Instruct-2507, Method Category=Indexing-enhanced RAG2026.05 | 20.7 |