Related papers: Calibrating Language Models with Adaptive Temperature Scaling

Calibrating Language Models with Adaptive Temperature Scaling

URL: http://arxiv.org/abs/2409.19817v1
Date: Sun, 29 Sep 2024 22:54:31 GMT
Title: Calibrating Language Models with Adaptive Temperature Scaling
Authors: Johnathan Xie, Annie S. Chen, Yoonho Lee, Eric Mitchell, Chelsea Finn,
Abstract summary: We introduce Adaptive Temperature Scaling (ATS), a post-hoc calibration method that predicts a temperature scaling parameter for each token prediction. ATS improves calibration by over 10-50% across three downstream natural language evaluation benchmarks compared to prior calibration methods.
Score: 58.056023173579625
License: http://creativecommons.org/licenses/by/4.0/
Abstract: The effectiveness of large language models (LLMs) is not only measured by their ability to generate accurate outputs but also by their calibration-how well their confidence scores reflect the probability of their outputs being correct. While unsupervised pre-training has been shown to yield LLMs with well-calibrated conditional probabilities, recent studies have shown that after fine-tuning with reinforcement learning from human feedback (RLHF), the calibration of these models degrades significantly. In this work, we introduce Adaptive Temperature Scaling (ATS), a post-hoc calibration method that predicts a temperature scaling parameter for each token prediction. The predicted temperature values adapt based on token-level features and are fit over a standard supervised fine-tuning (SFT) dataset. The adaptive nature of ATS addresses the varying degrees of calibration shift that can occur after RLHF fine-tuning. ATS improves calibration by over 10-50% across three downstream natural language evaluation benchmarks compared to prior calibration methods and does not impede performance improvements from RLHF.

Related papers

Feature Clipping for Uncertainty Calibration [24.465567005078135]
Modern deep neural networks (DNNs) often suffer from overconfidence, leading to miscalibration. We propose a novel post-hoc calibration method called feature clipping (FC) to address this issue. FC involves clipping feature values to a specified threshold, effectively increasing entropy in high calibration error samples.
arXiv Detail & Related papers (2024-10-16T06:44:35Z)
C-TPT: Calibrated Test-Time Prompt Tuning for Vision-Language Models via Text Feature Dispersion [54.81141583427542]
In deep learning, test-time adaptation has gained attention as a method for model fine-tuning without the need for labeled data. This paper explores calibration during test-time prompt tuning by leveraging the inherent properties of CLIP. We present a novel method, Calibrated Test-time Prompt Tuning (C-TPT), for optimizing prompts during test-time with enhanced calibration.
arXiv Detail & Related papers (2024-03-21T04:08:29Z)
Delving into temperature scaling for adaptive conformal prediction [10.340903334800787]
Conformal prediction, as an emerging uncertainty qualification technique, constructs prediction sets that are guaranteed to contain the true label with pre-defined probability. We show that current confidence calibration methods (e.g., temperature scaling) normally lead to larger prediction sets in adaptive conformal prediction. We propose $Conformal$ $Temperature$ $Scaling$ (ConfTS), a variant of temperature scaling that aims to improve the efficiency of adaptive conformal prediction.
arXiv Detail & Related papers (2024-02-06T19:27:48Z)
A Close Look into the Calibration of Pre-trained Language Models [56.998539510508515]
Pre-trained language models (PLMs) may fail in giving reliable estimates of their predictive uncertainty. We study the dynamic change in PLMs' calibration performance in training. We extend two recently proposed learnable methods that directly collect data to train models to have reasonable confidence estimations.
arXiv Detail & Related papers (2022-10-31T21:31:07Z)
Sample-dependent Adaptive Temperature Scaling for Improved Calibration [95.7477042886242]
Post-hoc approach to compensate for neural networks being wrong is to perform temperature scaling. We propose to predict a different temperature value for each input, allowing us to adjust the mismatch between confidence and accuracy. We test our method on the ResNet50 and WideResNet28-10 architectures using the CIFAR10/100 and Tiny-ImageNet datasets.
arXiv Detail & Related papers (2022-07-13T14:13:49Z)
Revisiting Calibration for Question Answering [16.54743762235555]
We argue that the traditional evaluation of calibration does not reflect usefulness of the model confidence. We propose a new calibration metric, MacroCE, that better captures whether the model assigns low confidence to wrong predictions and high confidence to correct predictions.
arXiv Detail & Related papers (2022-05-25T05:49:56Z)
T-Cal: An optimal test for the calibration of predictive models [49.11538724574202]
We consider detecting mis-calibration of predictive models using a finite validation dataset as a hypothesis testing problem. detecting mis-calibration is only possible when the conditional probabilities of the classes are sufficiently smooth functions of the predictions. We propose T-Cal, a minimax test for calibration based on a de-biased plug-in estimator of the $ell$-Expected Error (ECE)
arXiv Detail & Related papers (2022-03-03T16:58:54Z)
Parameterized Temperature Scaling for Boosting the Expressive Power in Post-Hoc Uncertainty Calibration [57.568461777747515]
We introduce a novel calibration method, Parametrized Temperature Scaling (PTS) We demonstrate that the performance of accuracy-preserving state-of-the-art post-hoc calibrators is limited by their intrinsic expressive power. We show with extensive experiments that our novel accuracy-preserving approach consistently outperforms existing algorithms across a large number of model architectures, datasets and metrics.
arXiv Detail & Related papers (2021-02-24T10:18:30Z)

This list is automatically generated from the titles and abstracts of the papers in this site.