Token-level and verbalized confidence signals in MLLMs frequently misalign, and a monotone fusion framework plus order-preserving alignment improves calibration and failure prediction across models.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Instinct vs. Reflection: Unifying Token and Verbalized Confidence in Multimodal Large Models
Token-level and verbalized confidence signals in MLLMs frequently misalign, and a monotone fusion framework plus order-preserving alignment improves calibration and failure prediction across models.