English
Related papers

Related papers: How Alignment Routes: Localizing, Scaling, and Con…

200 papers

Induction head mechanism is a part of the computational circuits for in-context learning (ICL) that enable large language models (LLMs) to adapt to new tasks without fine-tuning. Most existing work explains the training dynamics behind…

Computation and Language · Computer Science 2025-07-09 Shuo Wang , Issei Sato

Transformer-based pretrained large language models (PLM) such as BERT and GPT have achieved remarkable success in NLP tasks. However, PLMs are prone to encoding stereotypical biases. Although a burgeoning literature has emerged on…

Computation and Language · Computer Science 2024-06-18 Yi Yang , Hanyu Duan , Ahmed Abbasi , John P. Lalor , Kar Yan Tam

Conflict detection in policy languages is a solved problem -- as long as every rule condition is a crisp Boolean predicate. BDDs, SMT solvers, and NetKAT all exploit that assumption. But a growing class of routing and access-control systems…

Machine Learning · Computer Science 2026-03-20 Xunzhuo Liu , Hao Wu , Huamin Chen , Bowei He , Xue Liu

With the rapid development of Large Language Models (LLMs), aligning these models with human preferences and values is critical to ensuring ethical and safe applications. However, existing alignment techniques such as RLHF or DPO often…

Computation and Language · Computer Science 2025-08-19 Yang Zhang , Yu Yu , Bo Tang , Yu Zhu , Chuxiong Sun , Wenqiang Wei , Jie Hu , Zipeng Xie , Zhiyu Li , Feiyu Xiong , Edward Chung

Reasoning is a cognitive process of using evidence to reach a sound conclusion. The reasoning capability is essential for large language models (LLMs) to serve as the brain of the artificial general intelligence agent. Recent studies reveal…

Computation and Language · Computer Science 2023-09-06 Peiyi Wang , Lei Li , Liang Chen , Feifan Song , Binghuai Lin , Yunbo Cao , Tianyu Liu , Zhifang Sui

Standard fairness audits of foundation models quantify that a model is biased, but not where inside the network the bias resides. We propose a mechanistic fairness audit that combines projected residual-stream decomposition, zero-shot…

Computer Vision and Pattern Recognition · Computer Science 2026-03-13 Alaa Yasser , Kittipat Phunjanna , Marcos Escudero Viñolo , Catarina Barata , Jenny Benois-Pineau

This paper introduces an efficient and robust method for discovering interpretable circuits in large language models using discrete sparse autoencoders. Our approach addresses key limitations of existing techniques, namely computational…

Computation and Language · Computer Science 2024-05-22 Charles O'Neill , Thang Bui

Large language models internalize a structural trade-off between truthfulness and obsequious flattery, emerging from reward optimization that conflates helpfulness with polite submission. This latent bias, known as sycophancy, manifests as…

Computation and Language · Computer Science 2026-05-19 Sanskar Pandey , Ruhaan Chopra , Angkul Puniya , Sohom Pal

Can large language models introspect, that is, accurately detect perturbations to their own internal states? We systematically investigate this question using activation steering in Meta-Llama-3.1-8B-Instruct. First, we show that the binary…

Artificial Intelligence · Computer Science 2026-03-03 Ely Hahami , Ishaan Sinha , Lavik Jain , Josh Kaplan , Jon Hahami

The attention interaction matrix $QK^{\top}$ contains two entangled computations: a skew-symmetric component that redistributes information between positions (routing) and a symmetric component that scales mutual relevance (filtering). We…

Machine Learning · Computer Science 2026-05-20 Shafayeth Jamil , Rehan Kapadia

Current safety alignment for large language models(LLMs) continues to present vulnerabilities, given that adversarial prompting can effectively bypass their safety measures.Our investigation shows that these safety mechanisms predominantly…

Cryptography and Security · Computer Science 2025-08-28 Chao Huang , Zefeng Zhang , Juewei Yue , Quangang Li , Chuang Zhang , Tingwen Liu

RLHF-style alignment trains language models to refuse unsafe requests, but how much operational margin does this refusal rest on? We introduce the refusal-affirmation logit gap: the difference between the top refusal-token logit and the top…

Cryptography and Security · Computer Science 2026-05-05 Tung-Ling Li , Hongliang Liu

While modern LLMs are aligned to refuse harmful requests, it is essential to understand the underlying mechanistic basis of this refusal behavior for model safety analysis. For example, steering-based jailbreak attacks exploit this by…

Artificial Intelligence · Computer Science 2026-05-28 Su-Hyeon Kim , Hyundong Jin , Yejin Lee , Yo-Sub Han

Contextual priming, where earlier stimuli covertly bias later judgments, offers an unexplored attack surface for large language models (LLMs). We uncover a contextual priming vulnerability in which the previous response in the dialogue can…

Computation and Language · Computer Science 2025-11-24 Ziqi Miao , Lijun Li , Yuan Xiong , Zhenhua Liu , Pengyu Zhu , Jing Shao

Reinforcement Learning frameworks, particularly those utilizing human annotations, have become an increasingly popular method for preference fine-tuning, where the outputs of a language model are tuned to match a certain set of behavioral…

Machine Learning · Computer Science 2025-10-21 Archie Chaudhury

This study investigates the in-context learning capabilities of various decoder-only transformer-based language models with different model sizes and training data, including GPT2, SmolLM2, OpenELM, TinyLlama, Stable LM, and Gemma 2. We…

Computation and Language · Computer Science 2025-02-24 Yen-Che Hsiao , Abhishek Dutta

Vision-Language-Action (VLA) models have demonstrated strong potential for predicting semantic actions in navigation tasks, demonstrating the ability to reason over complex linguistic instructions and visual contexts. However, they are…

Despite significant advances in alignment techniques, we demonstrate that state-of-the-art language models remain vulnerable to carefully crafted conversational scenarios that can induce various forms of misalignment without explicit…

Computation and Language · Computer Science 2025-08-07 Siddhant Panpatil , Hiskias Dingeto , Haon Park

Aligned LLMs are secure, capable of recognizing and refusing to answer malicious questions. However, the role of internal parameters in maintaining such security is not well understood yet, further these models can be vulnerable to security…

Cryptography and Security · Computer Science 2025-04-08 Shen Li , Liuyi Yao , Lan Zhang , Yaliang Li

Masked diffusion language models (MDLMs) generate text via iterative masked-token denoising, enabling mask-parallel decoding and distinct controllability and efficiency tradeoffs from autoregressive LLMs. Yet, efficient representation-level…

Computation and Language · Computer Science 2026-03-31 Adi Shnaidman , Erin Feiglin , Osher Yaari , Efrat Mentel , Amit Levi , Raz Lapid
‹ Prev 1 4 5 6 7 8 10 Next ›