8 papers
Knowledge-Verified Emergent Deception in LLM Agents Under Conflicting Incentives
Zheyuan Liu, Weiliang Zhao, Xiangchi Yuan +3
Large language models are increasingly deployed as autonomous agents serving users on behalf of companies, placing them in settings where user and deployer interests can conflict.…
MTMCS-Bench: Evaluating Contextual Safety of Multimodal Large Language Models in Multi-Turn Dialogues
Zheyuan Liu, Dongwhi Kim, Yixin Wan +4
Multimodal large language models (MLLMs) are increasingly deployed as assistants that interact through text and images, making it crucial to evaluate contextual safety when risk de…
What Makes a Good Curriculum? Disentangling the Effects of Data Ordering on LLM Mathematical Reasoning
Yaning Jia, Chunhui Zhang, Xingjian Diao +4
Curriculum learning (CL) - ordering training data from easy to hard - has become a popular strategy for improving reasoning in large language models (LLMs). Yet prior work employs…
Steering Multimodal Large Language Models Decoding for Context-Aware Safety
Zheyuan Liu, Zhangchen Xu, Guangyao Dou +4
Multimodal Large Language Models (MLLMs) are increasingly deployed in real-world applications, yet their ability to make context-aware safety decisions remains limited. Existing me…
Growing Through Experience: Scaling Episodic Grounding in Language Models
Chunhui Zhang, Sirui, Wang +3
Language models (LMs) require robust episodic grounding-the capacity to learn from and apply past experiences-to excel at physical planning tasks. Current episodic grounding approa…
Superficial Self-Improved Reasoners Benefit from Model Merging
Xiangchi Yuan, Chunhui Zhang, Zheyuan Liu +4
As scaled language models (LMs) approach human-level reasoning capabilities, self-improvement emerges as a solution to synthesizing high-quality data corpus. While previous researc…
Modality-Aware Neuron Pruning for Unlearning in Multimodal Large Language Models
Zheyuan Liu, Guangyao Dou, Xiangchi Yuan +3
Generative models such as Large Language Models (LLMs) and Multimodal Large Language Models (MLLMs) trained on massive datasets can lead them to memorize and inadvertently reveal s…
AmoebaLLM: Constructing Any-Shape Large Language Models for Efficient and Instant Deployment
Yonggan Fu, Zhongzhi Yu, Junwei Li +6
Motivated by the transformative capabilities of large language models (LLMs) across various natural language tasks, there has been a growing demand to deploy these models effective…