The Metacognitive Probe identifies large within-model gaps in LLM confidence behavior, including a 47-point dissociation in Gemini 2.5 Flash between strong task calibration and weak difficulty prediction.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.AI 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
The Metacognitive Probe: Five Behavioural Calibration Diagnostics for LLMs
The Metacognitive Probe identifies large within-model gaps in LLM confidence behavior, including a 47-point dissociation in Gemini 2.5 Flash between strong task calibration and weak difficulty prediction.