Question Answering on NQ (F1)
67.8F1AGAR
Evaluation Results
| Method | Links | |
|---|---|---|
| AGAR2026.06 | 67.8 | |
| PANINIRetrieval Category=Proposed Method2026.02 | 67.44 | |
| VAR2026.06 | 66.9 | |
| Random2026.06 | 63 | |
| Uniform2026.06 | 62 | |
| AGLA2026.06 | 62 | |
| AGARBackbone=GLM-4.1V-9B-Thinking2026.06 | 61.8 | |
| Qwen3-Embedding (8B)Retrieval Category=Dense Retrieval, Reranker=Qwen3-Embedding (8B) + reranker2026.02 | 61.4 | |
| HippoRAG 2Retrieval Category=Structure-Augmented RAG2026.02 | 60 | |
| NV-Embed-v2 (7B)Retrieval Category=Dense Retrieval2026.02 | 59.9 | |
| Plain VQA2026.06 | 59.3 | |
| Qwen3-Embedding (8B)Retrieval Category=Dense Retrieval, Reranker=None2026.02 | 59.1 | |
| VARBackbone=GLM-4.1V-9B-Thinking2026.06 | 58.6 | |
| BM25Retrieval Category=Sparse Retrieval, Reranker=BM25 + reranker2026.02 | 58.2 | |
| CGR2026.06 | 56.2 | |
| BM25Retrieval Category=Sparse Retrieval, Reranker=None2026.02 | 56 | |
| AGLABackbone=GLM-4.1V-9B-Thinking2026.06 | 55.9 | |
| GraphRAGRetrieval Category=Structure-Augmented RAG2026.02 | 55.5 | |
| UniformBackbone=GLM-4.1V-9B-Thinking2026.06 | 55.4 | |
| Plain VQABackbone=GLM-4.1V-9B-Thinking2026.06 | 54.8 | |
| RandomBackbone=GLM-4.1V-9B-Thinking2026.06 | 54.7 | |
| RAPTORRetrieval Category=Structure-Augmented RAG2026.02 | 54.5 | |
| TIPSBackbone=Qwen-2.5-7B Instruct2026.03 | 53.22 | |
| NoneRetrieval Category=No Retrieval2026.02 | 52.7 | |
| HippoRAGRetrieval Category=Structure-Augmented RAG2026.02 | 52.2 | |
| TIPSBackbone=Qwen-2.5-3B Instruct2026.03 | 51.77 | |
| PPOBackbone=Qwen-2.5-3B Instruct2026.03 | 51.71 | |
| MT-PPOBackbone=Qwen-2.5-7B Instruct2026.03 | 50.6 | |
| PPOBackbone=Qwen-2.5-7B Instruct2026.03 | 50.53 | |
| VEA-Con2026.06 | 50.5 | |
| VEA-Br2026.06 | 49.4 | |
| MT-GRPO*Backbone=Qwen-2.5-7B Instruct2026.03 | 48.1 | |
| Search-R1Retrieval Category=Agentic Systems2026.02 | 47.9 | |
| GRPOBackbone=Qwen-2.5-7B Instruct2026.03 | 47.85 | |
| MT-PPOBackbone=Qwen-2.5-3B Instruct2026.03 | 47.75 | |
| GRPOBackbone=Qwen-2.5-3B Instruct2026.03 | 47.4 | |
| CGRBackbone=GLM-4.1V-9B-Thinking2026.06 | 47.4 | |
| VEA-ConBackbone=GLM-4.1V-9B-Thinking2026.06 | 47.3 | |
| MT-GRPO*Backbone=Qwen-2.5-3B Instruct2026.03 | 45.77 | |
| VEA-BrBackbone=GLM-4.1V-9B-Thinking2026.06 | 43.9 | |
| IRCoTRetrieval Category=Agentic Systems2026.02 | 26.4 | |
| LightRAGRetrieval Category=Structure-Augmented RAG2026.02 | 15.4 |