English
Related papers

Related papers: Semantic Gravity Wells: Why Negative Constraints B…

200 papers

The limits of open-ended generative models are unclear, yet increasingly important. What causes them to succeed and what causes them to fail? In this paper, we take a prompt-centric approach to analyzing and bounding the abilities of…

Computation and Language · Computer Science 2023-02-21 Albert Lu , Hongxin Zhang , Yanzhe Zhang , Xuezhi Wang , Diyi Yang

Despite the huge progress in myriad generation tasks, pretrained language models (LMs) such as GPT2 still tend to generate repetitive texts with maximization-based decoding algorithms for open-ended generation. We attribute their…

Computation and Language · Computer Science 2023-07-06 Jian Guan , Minlie Huang

There has been recent interest in whether large language models (LLMs) can introspect about their own internal states. Such abilities would make LLMs more interpretable, and also validate the use of standard introspective methods in…

Computation and Language · Computer Science 2025-09-25 Siyuan Song , Jennifer Hu , Kyle Mahowald

Small Language Models (SLMs) are increasingly being deployed in resource-constrained environments, yet their behavioral robustness to data contamination during instruction tuning remains poorly understood. We systematically investigate the…

Computation and Language · Computer Science 2025-11-11 Nicy Scaria , Silvester John Joseph Kennedy , Deepak Subramani

Fine-tuning language models on narrowly harmful data causes emergent misalignment (EM) -- behavioral failures extending far beyond training distributions. Recent work demonstrates compartmentalization of misalignment behind contextual…

Computation and Language · Computer Science 2026-03-06 Rohan Saxena

Numerous works are proposed to align large language models (LLMs) with human intents to better fulfill instructions, ensuring they are trustful and helpful. Nevertheless, some human instructions are often malicious or misleading and…

Computation and Language · Computer Science 2024-03-08 Rui Wang , Hongru Wang , Fei Mi , Yi Chen , Boyang Xue , Kam-Fai Wong , Ruifeng Xu

This paper discusses the problem of weakly supervised classification, in which instances are given weak labels that are produced by some label-corruption process. The goal is to derive conditions under which loss functions for weak-label…

Machine Learning · Statistics 2021-06-14 Shuhei M. Yoshida , Takashi Takenouchi , Masashi Sugiyama

Unsupervised on-the-fly back-translation, in conjunction with multilingual pretraining, is the dominant method for unsupervised neural machine translation. Theoretically, however, the method should not work in general. We therefore conduct…

Computation and Language · Computer Science 2024-03-28 Nicolas Guerin , Shane Steinert-Threlkeld , Emmanuel Chemla

Can in-context learning (ICL) override pre-trained label semantics, or does it merely refine an existing semantic backbone? We address this question by treating LLMs as prompt-induced classifiers and contrasting their behavior under…

Computation and Language · Computer Science 2025-11-27 Anantha Padmanaban Krishna Kumar

Large language models (LLMs) have a surprising failure: when trained on "A has a feature B", they do not generalize to "B is a feature of A", which is termed the Reversal Curse. Even when training with trillions of tokens this issue still…

Computation and Language · Computer Science 2024-05-09 Olga Golovneva , Zeyuan Allen-Zhu , Jason Weston , Sainbayar Sukhbaatar

Instruction-tuned Large Language Models (LLMs) have exhibited impressive language understanding and the capacity to generate responses that follow specific prompts. However, due to the computational demands associated with training these…

Computation and Language · Computer Science 2024-03-26 Yida Mu , Ben P. Wu , William Thorne , Ambrose Robinson , Nikolaos Aletras , Carolina Scarton , Kalina Bontcheva , Xingyi Song

Contrastive learning produces coherent semantic feature embeddings by encouraging positive samples to cluster closely while separating negative samples. However, existing contrastive learning methods lack principled guarantees on coverage…

Machine Learning · Computer Science 2026-03-30 Yahya Alkhatib , Wee Peng Tay

Safety alignment mechanism are essential for preventing large language models (LLMs) from generating harmful information or unethical content. However, cleverly crafted prompts can bypass these safety measures without accessing the model's…

Computation and Language · Computer Science 2025-01-31 Sunbowen Lee , Shiwen Ni , Chi Wei , Shuaimin Li , Liyang Fan , Ahmadreza Argha , Hamid Alinejad-Rokny , Ruifeng Xu , Yicheng Gong , Min Yang

Large language models (LLMs) are increasingly deployed with hierarchical instruction schemes, where certain instructions (e.g., system-level directives) are expected to take precedence over others (e.g., user messages). Yet, we lack a…

Computation and Language · Computer Science 2026-03-23 Yilin Geng , Haonan Li , Honglin Mu , Xudong Han , Timothy Baldwin , Omri Abend , Eduard Hovy , Lea Frermann

The role of semantics in zero-shot learning is considered. The effectiveness of previous approaches is analyzed according to the form of supervision provided. While some learn semantics independently, others only supervise the semantic…

Computer Vision and Pattern Recognition · Computer Science 2017-04-12 Pedro Morgado , Nuno Vasconcelos

We examine the ability of large language models (LLMs) to generate salient (interesting) negative statements about real-world entities; an emerging research topic of the last few years. We probe the LLMs using zero- and k-shot unconstrained…

Computation and Language · Computer Science 2023-09-22 Hiba Arnaout , Simon Razniewski

Instruction following aims to align Large Language Models (LLMs) with human intent by specifying explicit constraints on how tasks should be performed. However, we reveal a counterintuitive phenomenon: instruction following can…

Computation and Language · Computer Science 2026-01-30 Yunjia Qi , Hao Peng , Xintong Shi , Amy Xin , Xiaozhi Wang , Bin Xu , Lei Hou , Juanzi Li

Sentence compression reduces the length of text by removing non-essential content while preserving important facts and grammaticality. Unsupervised objective driven methods for sentence compression can be used to create customized models…

Computation and Language · Computer Science 2022-05-18 Demian Gholipour Ghalandari , Chris Hokamp , Georgiana Ifrim

Although the word-popularity based negative sampler has shown superb performance in the skip-gram model, the theoretical motivation behind oversampling popular (non-observed) words as negative samples is still not well understood. In this…

Machine Learning · Computer Science 2018-06-27 Long Chen , Fajie Yuan , Joemon M. Jose , Weinan Zhang

Reinforcement learning is hard in general. Yet, in many specific environments, learning is easy. What makes learning easy in one environment, but difficult in another? We address this question by proposing a simple measure of…

Machine Learning · Computer Science 2021-10-08 David Abel , Cameron Allen , Dilip Arumugam , D. Ellis Hershkowitz , Michael L. Littman , Lawson L. S. Wong