Multimodal Report Generation on OWID High-Density Multimodal Reports
49.86OrganizationHuman Gold-Standard (Ref)
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Human Gold-Standard (Ref)2026.04 | 49.86 | 50 | 50 | 50 | 50 | 49.97 | |
| CogGen (Ours)Backbone=GPT-4.1, Search LLM=GPT-4.1-Mini, Retrieval Tool=Tavily Search, Temperature=0.52026.04 | 49.72 | 58.13 | 50.42 | 48.06 | 43.26 | 49.92 | |
| WriteHereBackbone=GPT-4.1, Retrieval Tool=Tavily Search2026.04 | 49.12 | 55.03 | 49.36 | 38.46 | 33.12 | 45.02 | |
| STORMBackbone=GPT-4.1, Retrieval Tool=Tavily Search2026.04 | 42.53 | 44.43 | 39.86 | 16.75 | 16.67 | 32.05 | |
| Co-STORMBackbone=GPT-4.1, Retrieval Tool=Tavily Search2026.04 | 41.32 | 42.61 | 42.81 | 17.94 | 16.67 | 32.27 | |
| Multimodal DeepResearcherBackbone=GPT-4.1, Retrieval Tool=Tavily Search2026.04 | 37.68 | 42.93 | 35.08 | 18.19 | 17 | 30.18 |