Entity · technique

future confidence distillation

techniqueactiveprovisionalfuture-confidence-distillation-163c37f8·1 events·first seen 7h ago

Aliases: future confidence distillation

Co-occurring entities

Feeling-of-Knowing Judgement-of-Learning Future Confidence Distillation in Large Language Models

More like this (12)

Future Confidence Distillation in Large Language Models ensemble distillation Learning from the Self-future: On-policy Self-distillation for dLLMs on-policy distillation Weak-to-Strong Distillation Generalized Distillation Model Distillation knowledge distillation token-wise self-certainty grounded future mathematical generation Predicting Future Behaviors in Reasoning Models Enables Better Steering Rubric-Conditioned Self-Distillation

Recent events (1)

5arXiv · cs.CL·7h ago·source ↗

Future Confidence Distillation enables pre-solution confidence estimation in LLMs

Researchers introduce 'future confidence distillation,' a method that trains predictors on pre-solution hidden representations using post-solution correctness probes as teacher signals, enabling reliable confidence estimation before answer generation completes. The paper compares pre-solution Feeling-of-Knowing (FOK) and post-solution Judgement-of-Learning (JOL) confidence estimates across frontier and open-source LLMs, finding post-solution estimates are better calibrated and that linear probes recover richer confidence information than models verbalize. Distilled predictors recover much of the post-solution calibration improvement while remaining sample-efficient and domain-transferable. This has practical implications for confidence-aware systems involving retrieval, tool use, and adaptive computation.

Evaluation and Benchmarking Agent and Tool Ecosystem Feeling-of-Knowing Judgement-of-Learning Future Confidence Distillation in Large Language Models +1 more