Multimodal Question Answering on DesignQA
95Retrieval F1 (BoW)GPT-5-MCERF-Hybrid
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| GPT-5-MCERF-HybridCategory=Proposed Framework2026.01 | 95 | 55 | — | — | — | — | |
| GPT-5-MCERF-MainCategory=Proposed Framework2026.01 | 93 | 56 | 63 | 84 | 77 | 75 | |
| GPT-5-MCERF-HighReasoningCategory=Proposed Framework2026.01 | 92 | 51 | 64 | 85 | 80 | 81 | |
| GPT-4o-AllRulesCategory=AllRules Models2026.01 | 88 | 42 | 54 | 73 | 83 | 94 | |
| GPT-4-AllRulesCategory=AllRules Models2026.01 | 75 | 30 | 47 | 63 | 53 | 56 | |
| GPT-5-MCERF-SelfConsistencyCategory=Proposed Framework2026.01 | 71 | 56 | 56 | 82 | 75 | 75 | |
| GPT-4o-MCERF-MainCategory=Proposed Framework2026.01 | 61 | 42 | 54 | 74 | 75 | 75 | |
| GPT-4o-RAGCategory=RAG Models2026.01 | 19 | 38 | 53 | 71 | 68 | 75 | |
| GPT-4-RAGCategory=RAG Models2026.01 | 18 | 36 | 42 | 53 | 30 | 54 | |
| Claude-Opus-RAGCategory=RAG Models2026.01 | 17 | 29 | 42 | 51 | 51 | 88 | |
| LLaVA-1.5-RAGCategory=RAG Models2026.01 | 11 | 28 | 39 | 48 | 41 | 44 | |
| Baseline NaiveCategory=Baseline2026.01 | 8 | 14 | 36 | 50 | 50 | 50 | |
| Gemini-1.0-RAGCategory=RAG Models2026.01 | 0 | 28 | 49 | 55 | 53 | 88 | |
| GPT-5-MCERF-Vision2TextCategory=Proposed Framework2026.01 | — | — | 63 | 81 | 82 | 94 |