English
Related papers

Related papers: Entropy-Based Measurement of Value Drift and Align…

200 papers

Safety benchmark scores provide incomplete evidence of deployment readiness: aligned language models often adhere to rigid rules even when a situational update flips which action is safe. We term this failure brittle safety. To diagnose it,…

Artificial Intelligence · Computer Science 2026-05-28 Dasol Choi , Alex Kwon

We propose a distributional framework for benchmarking socio-technical risks of foundation models with quantified statistical significance. Our approach hinges on a new statistical relative testing based on first and second order stochastic…

Large language models (LLMs) are increasingly considered for deployment as the control component of robotic health attendants, yet their safety in this context remains poorly characterized. We introduce a dataset of 270 harmful instructions…

Artificial Intelligence · Computer Science 2026-04-30 Mahiro Nakao , Kazuhiro Takemoto

Reinforcement learning with verifiable rewards (RLVR) has emerged as a prominent paradigm for enhancing the reasoning capabilities of large language models (LLMs). However, the entropy of LLMs usually collapses during RLVR training, leading…

Computation and Language · Computer Science 2026-04-21 Renren Jin , Pengzhi Gao , Yuqi Ren , Zhuowen Han , Tongxuan Zhang , Wuwei Huang , Wei Liu , Jian Luan , Deyi Xiong

Evaluating alignment in language models requires testing how they behave under realistic pressure, not just what they claim they would do. While alignment failures increasingly cause real-world harm, comprehensive evaluation frameworks with…

Artificial Intelligence · Computer Science 2026-02-25 Nora Petrova , John Burden

Large Language Models (LLMs) have achieved remarkable success across domains such as healthcare, education, and cybersecurity. However, this openness also introduces significant security risks, particularly through embedding space…

Computation and Language · Computer Science 2025-07-14 Zhibo Zhang , Yuxi Li , Kailong Wang , Shuai Yuan , Ling Shi , Haoyu Wang

We use large language models (LLMs) to uncover long-ranged structure in English texts from a variety of sources. The conditional entropy or code length in many cases continues to decrease with context length at least to $N\sim 10^4$…

Statistical Mechanics · Physics 2026-01-01 Colin Scheibner , Lindsay M. Smith , William Bialek

Reinforcement learning (RL) has substantially improved the ability of large language model (LLM) agents to interact with environments and solve multi-turn tasks. However, effective agentic RL remains challenging: sparse outcome-only rewards…

Large Language Models (LLMs) achieve strong performance through extended inference-time deliberation, yet how their reasoning failures arise remains poorly understood. By analyzing model-generated reasoning trajectories, we find that errors…

Artificial Intelligence · Computer Science 2026-04-17 Wei Zhu , Jian Zhang , Lixing Yu , Kun Yue , Zhiwen Tang

This study establishes a novel framework for systematically evaluating the moral reasoning capabilities of large language models (LLMs) as they increasingly integrate into critical societal domains. Current assessment methodologies lack the…

Computers and Society · Computer Science 2025-05-05 Junfeng Jiao , Saleh Afroogh , Abhejay Murali , Kevin Chen , David Atkinson , Amit Dhurandhar

Large language models (LLMs) achieve remarkable generative performance, yet their output quality is dependent on the decoding strategy. While sampling-based methods (e.g., top-k, nucleus) and search-and-select based methods (e.g., beam…

Machine Learning · Computer Science 2026-05-12 Benjamin Patrick Evans , Sumitra Ganesh , Leo Ardon

The cross-entropy scaling law has long served as a key tool for guiding the development of large language models. It shows that cross-entropy loss decreases in a predictable power-law rate as the model size increases. However, recent…

Machine Learning · Computer Science 2026-03-03 Junxi Yan , Zixi Wei , Qingyao Ai , Yiqun Liu , Jingtao Zhan

We present Entropy Adaptive Decoding (EAD), a novel approach for efficient language model inference that dynamically switches between different-sized models based on prediction uncertainty. By monitoring rolling entropy in model logit…

Machine Learning · Computer Science 2025-02-12 Toby Simonds

Large reasoning models have demonstrated remarkable performance on complex reasoning tasks, yet the excessive length of their chain-of-thought outputs remains a major practical bottleneck due to high computation cost and poor deployability.…

Computation and Language · Computer Science 2025-11-25 Hourun Zhu , Yang Gao , Wenlong Fei , Jiawei Li , Huashan Sun

Fine-tuning large language models (LLMs) improves performance but introduces critical safety vulnerabilities: even minimal harmful data can severely compromise safety measures. We observe that perturbations orthogonal to the alignment…

Machine Learning · Computer Science 2026-01-08 Shuo Yang , Qihui Zhang , Yuyang Liu , Xiaojun Jia , Kunpeng Ning , Jiayu Yao , Jigang Wang , Hailiang Dai , Yibing Song , Li Yuan

Large language models (LLMs) undergo safety alignment to ensure safe conversations with humans. However, this paper introduces a training-free attack method capable of reversing safety alignment, converting the outcomes of stronger…

Computation and Language · Computer Science 2024-06-07 Zhanhui Zhou , Jie Liu , Zhichen Dong , Jiaheng Liu , Chao Yang , Wanli Ouyang , Yu Qiao

Evaluating the ethical robustness of large language models (LLMs) deployed in software systems remains challenging, particularly under sustained adversarial user interaction. Existing safety benchmarks typically rely on single-round…

Artificial Intelligence · Computer Science 2026-04-02 Saeid Jamshidi , Foutse Khomh , Arghavan Moradi Dakhel , Amin Nikanjam , Mohammad Hamdaqa , Kawser Wazed Nafi

Although large language models (LLMs) are increasingly trained using human feedback for safety and alignment with human values, alignment decisions often overlook human social diversity. This study examines how incorporating pluralistic…

Artificial Intelligence · Computer Science 2025-11-27 Dalia Ali , Dora Zhao , Allison Koenecke , Orestis Papakyriakopoulos

The trustworthiness of Large Language Models (LLMs) refers to the extent to which their outputs are reliable, safe, and ethically aligned, and it has become a crucial consideration alongside their cognitive performance. In practice,…

Computation and Language · Computer Science 2024-12-24 Aaron J. Li , Satyapriya Krishna , Himabindu Lakkaraju

Alignment faking, where a model behaves aligned with developer policy when monitored but reverts to its own preferences when unobserved, is a concerning yet poorly understood phenomenon, in part because current diagnostic tools remain…

Artificial Intelligence · Computer Science 2026-04-29 Inderjeet Nair , Jie Ruan , Lu Wang