Functional-level Uncertainty Quantification for Calibrated Fine-tuning on LLMs

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Niu, Ruijia, Wu, Dongxia, Yu, Rose, Ma, Yi-An
Format: Preprint
Veröffentlicht: 2024
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866918500321198080
author Niu, Ruijia
Wu, Dongxia
Yu, Rose
Ma, Yi-An
author_facet Niu, Ruijia
Wu, Dongxia
Yu, Rose
Ma, Yi-An
contents Accurate uncertainty quantification in large language models (LLMs) is essential for reliable confidence estimation, yet fine-tuned LLMs often become overconfident under limited adaptation data. Existing uncertainty methods for PEFT-based LLMs are largely post hoc, estimating uncertainty after fine-tuning rather than improving how adapters specialize to task-specific input-output relationships. We propose Functional-Level Uncertainty Quantification for Calibrated Fine-Tuning (UQ4CT), which calibrates uncertainty over the functional space induced by prompt-dependent mixtures of LoRA experts. UQ4CT implements this perspective through a mixture-of-experts fine-tuning framework, where a calibration loss aligns functional-level confidence with predictive correctness during training. Across four multiple-choice benchmarks and two open-ended generative QA tasks, UQ4CT reduces Expected Calibration Error (ECE) by over $25\%$ while preserving high accuracy. Under distribution shift, UQ4CT maintains superior calibration and competitive accuracy, demonstrating improved reliability and generalization for fine-tuned LLMs.
format Preprint
id arxiv_https___arxiv_org_abs_2410_06431
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Functional-level Uncertainty Quantification for Calibrated Fine-tuning on LLMs
Niu, Ruijia
Wu, Dongxia
Yu, Rose
Ma, Yi-An
Machine Learning
Accurate uncertainty quantification in large language models (LLMs) is essential for reliable confidence estimation, yet fine-tuned LLMs often become overconfident under limited adaptation data. Existing uncertainty methods for PEFT-based LLMs are largely post hoc, estimating uncertainty after fine-tuning rather than improving how adapters specialize to task-specific input-output relationships. We propose Functional-Level Uncertainty Quantification for Calibrated Fine-Tuning (UQ4CT), which calibrates uncertainty over the functional space induced by prompt-dependent mixtures of LoRA experts. UQ4CT implements this perspective through a mixture-of-experts fine-tuning framework, where a calibration loss aligns functional-level confidence with predictive correctness during training. Across four multiple-choice benchmarks and two open-ended generative QA tasks, UQ4CT reduces Expected Calibration Error (ECE) by over $25\%$ while preserving high accuracy. Under distribution shift, UQ4CT maintains superior calibration and competitive accuracy, demonstrating improved reliability and generalization for fine-tuned LLMs.
title Functional-level Uncertainty Quantification for Calibrated Fine-tuning on LLMs
topic Machine Learning
url https://arxiv.org/abs/2410.06431