Agentic-MME is a process-verified benchmark for multimodal agentic capabilities featuring 418 tasks, dual-axis checkpoints, and an overthinking metric that reveals even the best models achieve only 56.3% accuracy overall.
In other words, the bottleneck is no longer whether the correct region was surfaced, but whether the model can interpret the surfaced visual cue itself
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.AI 1years
2026 1verdicts
ACCEPT 1representative citing papers
citing papers explorer
-
Agentic-MME: What Agentic Capability Really Brings to Multimodal Intelligence?
Agentic-MME is a process-verified benchmark for multimodal agentic capabilities featuring 418 tasks, dual-axis checkpoints, and an overthinking metric that reveals even the best models achieve only 56.3% accuracy overall.