Mllms
Untracked preview. This is not a tracked Signal Bureau entity — the page is generated on request from live coverage search, is not indexed, and carries no editorial curation. Tracked entities live on the Wire.
If Mllms is on your watch list: what changed in the coverage, what the reporting asserts, and what the crowd is pricing.
What’s changed
- Teaching MLLMs to Say No: Generalized Referring Expression Comprehension via Refusal Calibrated GRPO · arXiv cs.AI · 2026-08-06
- LongChart VQA: A Comprehensive Benchmark for MLLMs with Complex Multi-Chart Reasoning · ArXiv cs.AI · 2026-08-05
- EgoMonth: A Month-Level Egocentric Video Benchmark for Long-Term Spatiotemporal Memory · ArXiv cs.AI · 2026-08-15
- Diagram-MMU: A Multi-Modal Benchmark for Scientific Diagrams · ArXiv cs.AI · 2026-08-13
- Evidence-Grounded Trustworthy Multimodal Reasoning and Evaluation Benchmark in Complex Urban Scenes · arXiv cs.AI · 2026-08-12
- Science Edge Evaluation: SEE the Missing Step Toward Real Scientific Discovery · arXiv cs.AI · 2026-08-10
What the reporting says
It shows up alongside Computer Science (14), Large Language Models (14), Abstract (14), View Pdf Html (13), Aug (10).
Two steps out: AI (through Computer Science, 14 shared stories on the weaker link); Anthropic (through Computer Science, 14 shared stories on the weaker link); Google (through Large Language Models, 14 shared stories on the weaker link); Huggingface (through Computer Science, 14 shared stories on the weaker link).
Tracked since Jul 2, with coverage on 12 separate days — concentrated in Artificial Intelligence and AI Investing.