Question Answering on Medical
68.81GPT AccuracyGAM-RAG
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| GAM-RAGMemorization turns=22026.03 | 68.81 | 63.23 | — | |
| GAM-RAGMemorization turns=12026.03 | 66.72 | 61.92 | — | |
| GAM-RAGMemorization turns=02026.03 | 65.47 | 58.98 | — | |
| EHRAGGeneration Strategy=Retrieval-Augmented-Generation Methods2026.04 | 65.32 | — | — | |
| REMINDRAGMemorization turns=02026.03 | 64.25 | 56.22 | — | |
| LinearRAGMemorization turns=02026.03 | 63.72 | 54.92 | — | |
| LinearRAGGeneration Strategy=Retrieval-Augmented-Generation Methods2026.04 | 63.72 | — | — | |
| Vanilla RAGGeneration Strategy=Retrieval-Augmented-Generation Methods2026.04 | 61.68 | — | — | |
| nvidia/NV-Embed-v2Memorization turns=02026.03 | 61.44 | 56.27 | — | |
| PoGMemorization turns=02026.03 | 61.4 | 53.76 | — | |
| all-mpnet-base-v2Memorization turns=02026.03 | 60.96 | 55.3 | — | |
| HippoRAG2Memorization turns=02026.03 | 60.77 | 52.48 | — | |
| HippoRAG2Generation Strategy=Retrieval-Augmented-Generation Methods2026.04 | 60.77 | — | — | |
| GritLM / GritLM-7BMemorization turns=02026.03 | 58.83 | 54.21 | — | |
| Qwen3-8BGeneration Strategy=Direct Zero-shot LLM Inference2026.04 | 58.39 | — | — | |
| E2GraphRAGGeneration Strategy=Retrieval-Augmented-Generation Methods2026.04 | 58 | — | — | |
| GFM-RAGMemorization turns=02026.03 | 56.07 | 50.24 | — | |
| GFM-RAGGeneration Strategy=Retrieval-Augmented-Generation Methods2026.04 | 56.07 | — | — | |
| RAPTORGeneration Strategy=Retrieval-Augmented-Generation Methods2026.04 | 55.75 | — | — | |
| HippoRAGGeneration Strategy=Retrieval-Augmented-Generation Methods2026.04 | 55.04 | — | — | |
| LightRAGGeneration Strategy=Retrieval-Augmented-Generation Methods2026.04 | 54.36 | — | — | |
| KGPGeneration Strategy=Retrieval-Augmented-Generation Methods2026.04 | 54.22 | — | — | |
| DyG-RAGMemorization turns=02026.03 | 51.28 | 44.92 | — | |
| G-retrieverGeneration Strategy=Retrieval-Augmented-Generation Methods2026.04 | 50.36 | — | — | |
| GraphRAGGeneration Strategy=Retrieval-Augmented-Generation Methods2026.04 | 48.5 | — | — | |
| GPT-3.5-turboGeneration Strategy=Direct Zero-shot LLM Inference2026.04 | 45.6 | — | — | |
| Chain-of-Thought (CoT)Memorization turns=02026.03 | 42.52 | 34.5 | — | |
| Direct AnsweringMemorization turns=02026.03 | 42.1 | 32.16 | — | |
| GPT-4o-miniGeneration Strategy=Direct Zero-shot LLM Inference2026.04 | 42.1 | — | — | |
| llama-13BGeneration Strategy=Direct Zero-shot LLM Inference2026.04 | 28.86 | — | — | |
| llama-8BGeneration Strategy=Direct Zero-shot LLM Inference2026.04 | 27.31 | — | — | |
| BM25Inference LLM=Gemini-2.5-flash2025.08 | — | 37.8 | 22.2 | |
| BM25Inference LLM=GPT-4o-mini2025.08 | — | 43.6 | 25.9 | |
| HippoRAG2Inference LLM=Gemini-2.5-flash2025.08 | — | 58.2 | 27.8 | |
| HippoRAG2Inference LLM=GPT-4o-mini2025.08 | — | 59.4 | 28.1 | |
| IRCoTInference LLM=Gemini-2.5-flash2025.08 | — | 56.1 | 26.1 | |
| IRCoTInference LLM=GPT-4o-mini2025.08 | — | 52.9 | 25.1 | |
| NORInference LLM=Gemini-2.5-flash2025.08 | — | 46 | 23.1 | |
| NORInference LLM=GPT-4o-mini2025.08 | — | 44.2 | 19.3 | |
| RAPTORInference LLM=Gemini-2.5-flash2025.08 | — | 58.1 | 34.2 | |
| RAPTORInference LLM=GPT-4o-mini2025.08 | — | 63.7 | 36.7 | |
| StandardInference LLM=Gemini-2.5-flash2025.08 | — | 41.5 | 19.4 | |
| StandardInference LLM=GPT-4o-mini2025.08 | — | 59.9 | 27 | |
| T^2RAGInference LLM=Gemini-2.5-flash2025.08 | — | 61.4 | 36 | |
| T^2RAGInference LLM=GPT-4o-mini2025.08 | — | 60.4 | 33.5 |