English
Related papers

Related papers: Where does output diversity collapse in post-train…

200 papers

Self-play bootstraps LLM reasoning through an iterative Challenger-Solver loop: the Challenger is trained to generate questions that target the Solver's capabilities, and the Solver is optimized on the generated data to expand its reasoning…

Machine Learning · Computer Science 2026-02-17 Gengsheng Li , Jinghan He , Shijie Wang , Dan Zhang , Ruiqi Liu , Renrui Zhang , Zijun Yao , Junfeng Fang , Haiyun Guo , Jinqiao Wang

Whether machines can originate novel content has been debated for nearly two centuries, from Lovelace's assertion that no engine can "originate anything" to Turing's question of whether a machine can amplify ideas brought in from outside.…

Multiagent Systems · Computer Science 2026-05-19 Weiyi Kong , Shiyang Lai , Jinghua Piao , James Evans

Reinforcement learning with verifiable outcome rewards (RLVR) has effectively scaled up chain-of-thought (CoT) reasoning in large language models (LLMs). Yet, its efficacy in training vision-language model (VLM) agents for goal-directed…

Computer Vision and Pattern Recognition · Computer Science 2025-07-14 Tong Wei , Yijun Yang , Junliang Xing , Yuanchun Shi , Zongqing Lu , Deheng Ye

Chain of thought finetuning (cot-finetuning) aims to endow small language models (SLM) with reasoning ability to improve their performance towards specific tasks by allowing them to imitate the reasoning procedure of large language models…

Computation and Language · Computer Science 2025-08-08 Xiaoshu Chen , Sihang Zhou , Ke Liang , Xinwang Liu

On-Policy Self-Distillation (OPSD) has recently emerged as an alternative to Reinforcement Learning with Verifiable Rewards (RLVR), promising higher accuracy and shorter responses through token-level credit assignment from a self-teacher…

Artificial Intelligence · Computer Science 2026-05-08 Jaehoon Kim , Dongha Lee

Foundation models encode rich structural knowledge but often rely on post-training procedures to adapt their reasoning behavior to specific tasks. Popular approaches such as reinforcement learning with verifiable rewards (RLVR) and…

Machine Learning · Computer Science 2026-01-21 Dake Bu , Wei Huang , Andi Han , Atsushi Nitanda , Bo Xue , Qingfu Zhang , Hau-San Wong , Taiji Suzuki

Post-training alignment often reduces LLM diversity, leading to a phenomenon known as mode collapse. Unlike prior work that attributes this effect to algorithmic limitations, we identify a fundamental, pervasive data-level driver:…

Computation and Language · Computer Science 2025-10-13 Jiayi Zhang , Simon Yu , Derek Chong , Anthony Sicilia , Michael R. Tomz , Christopher D. Manning , Weiyan Shi

The notion of neural collapse refers to several emergent phenomena that have been empirically observed across various canonical classification problems. During the terminal phase of training a deep neural network, the feature embedding of…

Machine Learning · Computer Science 2023-04-05 Duc Anh Nguyen , Ron Levie , Julian Lienen , Gitta Kutyniok , Eyke Hüllermeier

With the remarkable generative capabilities of large language models (LLMs), using LLM-generated data to train downstream models has emerged as a promising approach to mitigate data scarcity in specific domains and reduce time-consuming…

Computation and Language · Computer Science 2025-06-26 Yuchang Zhu , Huazhen Zhong , Qunshu Lin , Haotong Wei , Xiaolong Sun , Zixuan Yu , Minghao Liu , Zibin Zheng , Liang Chen

Supervised Fine-Tuning (SFT) on long Chain-of-Thought (CoT) trajectories has become a pivotal phase in building large reasoning models. However, how CoT trajectories from different sources influence the generalization performance of models…

Computation and Language · Computer Science 2026-04-07 Zhaoyi Li , Xiangyu Xi , Zhengyu Chen , Wei Wang , Gangwei Jiang , Ranran Shen , Linqi Song , Ying Wei , Defu Lian

Recent work on enhancing the reasoning abilities of large language models (LLMs) has introduced explicit length control as a means of constraining computational cost while preserving accuracy. However, existing approaches rely on…

Computation and Language · Computer Science 2025-08-13 Hasan Abed Al Kader Hammoud , Kumail Alhamoud , Abed Hammoud , Elie Bou-Zeid , Marzyeh Ghassemi , Bernard Ghanem

System 2 reasoning is one of the defining characteristics of intelligence, which requires slow and logical thinking. Human conducts System 2 reasoning via the language of thoughts that organizes the reasoning process as a causal sequence of…

Computation and Language · Computer Science 2025-05-20 Chenxi Liu , Yongqiang Chen , Tongliang Liu , James Cheng , Bo Han , Kun Zhang

Inference-time computation has emerged as a promising scaling axis for improving large language model reasoning. However, despite yielding impressive performance, the optimal allocation of inference-time computation remains poorly…

Machine Learning · Computer Science 2026-01-12 Parsa Mirtaheri , Ezra Edelman , Samy Jelassi , Eran Malach , Enric Boix-Adsera

Large language models often generate homogeneous outputs, but whether this is problematic depends on the specific task. For objective math tasks, responses may vary in terms of problem-solving strategy but should maintain the same…

Computation and Language · Computer Science 2026-04-23 Shomik Jain , Jack Lanchantin , Maximilian Nickel , Candace Ross , Karen Ullrich , Ashia Wilson , Jamelle Watson-Daniels

Large Language Models (LLMs) often exhibit strong linguistic abilities while remaining unreliable on multi-step reasoning tasks, particularly when deployed without additional training or fine-tuning. In this work, we study inference-time…

Computation and Language · Computer Science 2026-03-24 Vinay Sharma , Manish Jain

Large language models (LLMs) have been routinely used to solve various tasks using step-by-step reasoning. However, the structure of intermediate reasoning steps, or thoughts, is rigid and unidirectional, such as chains, trees, or…

Artificial Intelligence · Computer Science 2024-12-30 Sijia Chen , Baochun Li

The class-wise training losses often diverge as a result of the various levels of intra-class and inter-class appearance variation, and we find that the diverging class-wise training losses cause the uncalibrated prediction with its…

Machine Learning · Computer Science 2023-06-21 Seungjin Jung , Seungmo Seo , Yonghyun Jeong , Jongwon Choi

When the substantive content of a request is rewritten, do large language models still answer in the format the original task asked for? We find that they often do not, even at temperature zero. On a 150-query evaluation over five compact…

Computation and Language · Computer Science 2026-05-12 Aofan Liu , Jingxiang Meng

The extraordinary capabilities of large language models (LLMs) such as ChatGPT and GPT-4 are in part unleashed by aligning them with reward models that are trained on human preferences, which are often represented as rankings of responses…

Machine Learning · Computer Science 2025-10-31 Ziang Song , Tianle Cai , Jason D. Lee , Weijie J. Su

Large language models (LLMs) can perform reasoning computations both internally within their latent space and externally by generating explicit token sequences like chains of thought. Significant progress in enhancing reasoning abilities…

Computation and Language · Computer Science 2025-04-16 Thilo Hagendorff , Sarah Fabi