10 papers
Localize-Then-Decide Guarantees for LLM Judgments
Xinyu Li, Yi Zhou, Guanqun Cao +3
Large language models (LLMs) are increasingly used as evaluators to assess output quality and preference alignment, yet providing reliable guarantees of agreement with human judgme…
Margin-Adaptive Confidence Ranking for Reliable LLM Judgement
Gaojie Jin, Yong Tao, Lijia Yu +1
Jung et al. (2025) introduce a hypothesis testing framework for guaranteeing agreement between large language models (LLMs) and human judgments, relying on the assumption that the…
OTora: A Unified Red Teaming Framework for Reasoning-Level Denial-of-Service in LLM Agents
Xinyu Li, Ronghui Mu, Lin Li +2
Large Language Models (LLMs) are increasingly deployed as autonomous agents that execute tool-augmented, multi-step tasks, where latency is a critical factor for real-world applica…
POT: Inducing Overthinking in LLMs via Black-Box Iterative Optimization
Xinyu Li, Tianjin Huang, Ronghui Mu +2
Recent advances in Chain-of-Thought (CoT) prompting have substantially enhanced the reasoning capabilities of large language models (LLMs), enabling sophisticated problem-solving t…
LOST: Low-rank and Sparse Pre-training for Large Language Models
Jiaxi Li, Lu Yin, Li Shen +6
While large language models (LLMs) have achieved remarkable performance across a wide range of tasks, their massive scale incurs prohibitive computational and memory costs for pre-…
Principal Eigenvalue Regularization for Improved Worst-Class Certified Robustness of Smoothed Classifiers
Gaojie Jin, Tianjin Huang, Ronghui Mu +1
Recent studies have identified a critical challenge in deep neural networks (DNNs) known as ``robust fairness", where models exhibit significant disparities in robust accuracy acro…
SPAM: Spike-Aware Adam with Momentum Reset for Stable LLM Training
Tianjin Huang, Ziquan Zhu, Gaojie Jin +3
Large Language Models (LLMs) have demonstrated exceptional performance across diverse tasks, yet their training remains highly resource-intensive and susceptible to critical challe…
GradientStabilizer:Fix the Norm, Not the Gradient
Tianjin Huang, Zhangyang Wang, Haotian Hu +10
Training instability in modern deep learning systems is frequently triggered by rare but extreme gradient-norm spikes, which can induce oversized parameter updates, corrupt optimiz…
Enhancing Robust Fairness via Confusional Spectral Regularization
Gaojie Jin, Sihao Wu, Jiaxu Liu +2
Recent research has highlighted a critical issue known as ``robust fairness", where robust accuracy varies significantly across different classes, undermining the reliability of de…
Composable Interventions for Language Models
Arinbjorn Kolbeinsson, Kyle O'Brien, Tianjin Huang +8
Test-time interventions for language models can enhance factual accuracy, mitigate harmful outputs, and improve model efficiency without costly retraining. But despite a flood of n…