paper-with-me

홈 › Papers

Optimizing Calibration by Gaining Aware of Prediction Correctness

2024-04-19 · Yuchi Liu, Lei Wang, Yuli Zou, James Zou, Liang Zheng

Model calibration aims to align confidence with prediction correctness. The Cross-Entropy (CE) loss is widely used for calibrator training, which enforces the model to increase confidence on the ground truth class. However, we find the CE loss has intrinsic limitations. For example, for a narrow misclassification, a calibrator trained by the CE loss often produces high confidence on the wrongly predicted class (e.g., a test sample is wrongly classified and its softmax score on the ground truth class is around 0.4), which is undesirable. In this paper, we propose a new post-hoc calibration objective derived from the aim of calibration. Intuitively, the proposed objective function asks that the calibrator decrease model confidence on wrongly predicted samples and increase confidence on correctly predicted samples. Because a sample itself has insufficient ability to indicate correctness, we use its transformed versions (e.g., rotated, greyscaled and color-jittered) during calibrator training. Trained on an in-distribution validation set and tested with isolated, individual test samples, our method achieves competitive calibration performance on both in-distribution and out-of-distribution test sets compared with the state of the art. Further, our analysis points out the difference between our method and commonly used objectives such as CE loss and mean square error loss, where the latters sometimes deviates from the calibration aim.

📄 PDF Abstract BibTeX arXiv:2404.13016

Code (1)

liuyvchi/correctness-aware-calibration 공식 구현 pytorch

Tasks

Prediction

Methods 이 논문이 사용한 방법론

SET Dynamic Sparse Training method where weight mask is updated randomly periodically
Softmax The Softmax output function transforms a previous layer's output into a vector of probabilities. It is commonly used for multiclass classification. Given an input vector $x$…
ALIGN In the ALIGN method, visual and language representations are jointly trained from noisy image alt-text data. The image and text encoders are learned via contrastive loss…

Similar Papers 제목 키워드 기반

CATTO: Balancing Preferences and Confidence in Language Models

2026-01-30 · Nisarg Parikh, Ananya Sai, Pannaga Shivaswamy, Kunjal Panchal 외 arxiv

Large language models (LLMs) often make accurate next token predictions but their confidence in these predictions can be poorly calibrated: high-confidence predictions are frequently wrong, and low-confidence predictions…

LLM-VeriPPA: Power, Performance, and Area Optimization aware Verilog Code Generation with Large Language Models

2025-09-10 · Kiran Thorat, Jiahui Zhao, Yaotian Liu, Amit Hasan 외 arxiv

Large Language Models (LLMs) are gaining prominence in various fields, thanks to their ability to generate high- quality content from human instructions. This paper delves into the field of chip design using LLMs, specif…

Code Generation

ORCE: Order-Aware Alignment of Verbalized Confidence in Large Language Models

2026-05-12 · Chen Li, Xiaoling Hu, Songzhu Zheng, Jiawei Zhou 외 arxiv

Large language models (LLMs) often produce answers with high certainty even when they are incorrect, making reliable confidence estimation essential for deployment in real-world scenarios. Verbalized confidence, where mo…

Reinforcement LearningAnswer Generation

Semantic-Aware Confidence Calibration for Automated Audio Captioning

2025-12-11 · Lucas Dunker, Sai Akshay Menta, Snigdha Mohana Addepalli, Venkata Krishna Rayalu Garapati arxiv

Automated audio captioning models frequently produce overconfident predictions regardless of semantic accuracy, limiting their reliability in deployment. This deficiency stems from two factors: evaluation metrics based o…

Semantic SimilarityAudio captioning

Know What You Know: Metacognitive Entropy Calibration for Verifiable RL Reasoning

2026-02-26 · Qiannian Zhao, Chen Yang, Jinhao Jing, Yunke Zhang 외 arxiv

Large reasoning models (LRMs) have emerged as a powerful paradigm for solving complex real-world tasks. In practice, these models are predominantly trained via Reinforcement Learning with Verifiable Rewards (RLVR), yet m…

Reinforcement LearningQuestion Answering