Future Confidence Distillation in Large Language Models
Quick Answer
This study introduces future confidence distillation for large language models (LLMs), enhancing confidence estimation by leveraging post-solution correctness probes.
Quick Take
This study introduces future confidence distillation for large language models (LLMs), enhancing confidence estimation by leveraging post-solution correctness probes. The method shows that post-solution confidence is better calibrated than pre-solution estimates, allowing for more reliable and efficient confidence predictions across datasets, significantly improving decision-making in confidence-aware systems.
Key Points
- Post-solution confidence is better calibrated than pre-solution estimates.
- Future confidence distillation uses hidden representations for efficient inference.
- Method shows high sample efficiency and transfers across datasets.
- Confidence-related information evolves during the answering process.
- Improved confidence estimation aids in retrieval and adaptive computation.
Paper Resources
📖 Reader Mode
~2 min readAbstract:Reliable confidence estimation is essential for deploying large language models (LLMs) in confidence-aware systems, where downstream decisions such as retrieval, tool use, and adaptive computation depend on accurately estimating answer reliability. Existing approaches, however, largely treat confidence as a property of completed responses, overlooking how confidence-related information evolves throughout the answering process. In this work, we investigate confidence from a temporal perspective by comparing pre-solution Feeling-of-Knowing (FOK) and post-solution Judgement-of-Learning (JOL) confidence estimates across frontier and open-source LLMs. We show that post-solution confidence is consistently better calibrated and more discriminative than pre-solution confidence, while linear probes trained on hidden representations recover substantially richer confidence-related information than models explicitly verbalise. Building on this observation, we introduce future confidence distillation, which trains predictors operating on pre-solution hidden representations using teacher confidence estimates produced by post-solution correctness probes. Despite requiring only pre-solution representations for inference, distilled predictors recover much of the calibration improvement achieved by post-solution confidence, remain highly sample efficient, and transfer across datasets within the same domain. Together, our findings demonstrate that confidence-related information evolves throughout the answering process and can be anticipated before answer generation is complete, enabling significantly more reliable yet low-cost confidence estimation.
| Comments: | 16 pages, 5 figures |
| Subjects: | Computation and Language (cs.CL); Artificial Intelligence (cs.AI) |
| ACM classes: | I.2.7 |
| Cite as: | arXiv:2607.07626 [cs.CL] |
| (or arXiv:2607.07626v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2607.07626 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Sahil Kale [view email]
[v1]
Wed, 8 Jul 2026 16:43:11 UTC (9,415 KB)
— Originally published at arxiv.org
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from arXiv cs.CL
See more →Quantifying Prior Dominance in Systems
The study introduces the Normalized Context Utilization (NCU) metric to evaluate Retrieval-Augmented Generation (RAG) systems, revealing that Small Language Models (SLMs) outperform larger models in factual extraction. The findings indicate that traditional scaling laws yield diminishing returns, with a commercial API frequently failing against adversarial evidence due to systemic confidence collapse.