LLMs struggle to associate epistemic markers with stable internal confidence levels across distributions, even under model-centric interpretations, while maintaining somewhat consistent marker rankings.
Can Large Language Models Faithfully Express Their Intrinsic Uncertainty in Words?
5 Pith papers cite this work. Polarity classification is still indexing.
5
Pith papers citing it
years
2026 5verdicts
UNVERDICTED 5representative citing papers
LMs systematically inflate expressed certainty during rewriting, affecting up to 75% of outputs with a 1.5-2x bias toward increasing rather than decreasing certainty, and the effect compounds over iterations.
A new framework quantifies faithful confidence expression in large reasoning models by comparing linguistic decisiveness to token probabilities, hidden states, and response consistency, revealing it as a persistent challenge.
BAG prompts LLMs to reason over K sampled responses for strategy selection in multi-turn ambiguous QA, improving accuracy and faithfulness to uncertainty over baselines across six models.
RLCM trains LLMs with a margin-enhanced process reward that widens the gap between correct and incorrect reasoning steps, improving calibration on math, code, logic, and science tasks without hurting accuracy.
citing papers explorer
-
Process Supervision of Confidence Margin for Calibrated LLM Reasoning
RLCM trains LLMs with a margin-enhanced process reward that widens the gap between correct and incorrect reasoning steps, improving calibration on math, code, logic, and science tasks without hurting accuracy.