A new five-level evaluation framework and a 702-task multimodal benchmark show that current multimodal LLMs rarely surpass task-specific specialists, so none reach the highest 'total synergy' level.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
On Path to Multimodal Generalist: General-Level and General-Bench
A new five-level evaluation framework and a 702-task multimodal benchmark show that current multimodal LLMs rarely surpass task-specific specialists, so none reach the highest 'total synergy' level.