5 papers
Provable Joint Decontamination for Benchmarking Multiple Large Language Models
Zhenlong Liu, Hao Zeng, Hongxin Wei
Benchmark data contamination has become a central challenge in LLM evaluation: when evaluation examples appear in the training data of one or more audited models, reported performa…
Detecting Distillation Data from Reasoning Models
Hengxiang Zhang, Hyeong Kyu Choi, Sharon Li +1
Reasoning distillation has emerged as a prevailing paradigm for transferring reasoning capabilities from large reasoning models to small language models. Yet, reasoning distillatio…
Unlocking the Pre-Trained Model as a Dual-Alignment Calibrator for Post-Trained LLMs
Beier Luo, Cheng Wang, Hongxin Wei +2
Post-training improves large language models (LLMs) but often worsens confidence calibration, leading to systematic overconfidence. Recent unsupervised post-hoc methods for post-tr…
Your Pre-trained LLM is Secretly an Unsupervised Confidence Calibrator
Beier Luo, Shuoyuan Wang, Sharon Li +1
Post-training of large language models is essential for adapting pre-trained language models (PLMs) to align with human preferences and downstream tasks. While PLMs typically exhib…
How Contaminated Is Your Benchmark? Quantifying Dataset Leakage in Large Language Models with Kernel Divergence
Hyeong Kyu Choi, Maxim Khanov, Hongxin Wei +1
Dataset contamination, where evaluation datasets overlap with pre-training corpora, inflates performance metrics and undermines the reliability of model evaluations. Measuring data…