12 papers
LaCache: Robust Semantic Caching for LLM Serving
Jiacheng Liang, Yuhui Wang, Tanqiu Jiang +1
Semantic caching, reusing responses to answer semantically similar queries, has seen growing adoption in LLM serving, offering faster responses and reduced costs. Yet existing sche…
Inference-Time Conformal Reasoning with Valid Factuality Control for Large Language Models
Ting Wang, Yuanjie Shi, Yan Yan +1
Large language models (LLMs) increasingly perform multi-step reasoning, where intermediate claims form implicit directed acyclic graphs whose node correctness is structurally condi…
Safeguarding LLM Agents against Long-Horizon Threats via Shadow Memory
Yuhui Wang, Tanqiu Jiang, Jiacheng Liang +2
As large language model (LLM)-powered agents are increasingly deployed to perform complex, real-world tasks, they face a growing class of attacks that exploit extended user-agent-e…
Routing-Aware Safety Alignment for Mixture-of-Experts Models
Jiacheng Liang, Yuhui Wang, Tanqiu Jiang +1
Mixture-of-Experts (MoE) language models introduce unique challenges for safety alignment due to their sparse routing mechanisms, which can enable degenerate optimization behaviors…
SafeGRPO: Self-Rewarded Multimodal Safety Alignment via Rule-Governed Policy Optimization
Xuankun Rong, Wenke Huang, Tingfeng Wang +3
Multimodal large language models (MLLMs) have demonstrated impressive reasoning and instruction-following capabilities, yet their expanded modality space introduces new composition…
Reasoning or Retrieval? A Study of Answer Attribution on Large Reasoning Models
Yuhui Wang, Changjiang Li, Guangke Chen +2
Large reasoning models (LRMs) exhibit unprecedented capabilities in solving complex problems through Chain-of-Thought (CoT) reasoning. However, recent studies reveal that their fin…
Self-Improving Model Steering
Rongyi Zhu, Yuhui Wang, Tanqiu Jiang +2
Model steering represents a powerful technique that dynamically aligns large language models (LLMs) with human preferences during inference. However, conventional model-steering me…
Dynamic Token Reweighting for Robust Vision-Language Models
Tanqiu Jiang, Jiacheng Liang, Rongyi Zhu +3
Large vision-language models (VLMs) are highly vulnerable to multimodal jailbreak attacks that exploit visual-textual interactions to bypass safety guardrails. In this paper, we pr…
Self-Destructive Language Model
Yuhui Wang, Rongyi Zhu, Ting Wang
Harmful fine-tuning attacks pose a major threat to the security of large language models (LLMs), allowing adversaries to compromise safety guardrails with minimal harmful data. Whi…
AutoRAN: Automated Hijacking of Safety Reasoning in Large Reasoning Models
Jiacheng Liang, Tanqiu Jiang, Yuhui Wang +3
This paper presents AutoRAN, the first framework to automate the hijacking of internal safety reasoning in large reasoning models (LRMs). At its core, AutoRAN pioneers an execution…
RAPID: Retrieval Augmented Training of Differentially Private Diffusion Models
Tanqiu Jiang, Changjiang Li, Fenglong Ma +1
Differentially private diffusion models (DPDMs) harness the remarkable generative capabilities of diffusion models while enforcing differential privacy (DP) for sensitive data. How…
GraphRAG under Fire
Jiacheng Liang, Yuhui Wang, Changjiang Li +4
GraphRAG advances retrieval-augmented generation (RAG) by structuring external knowledge as multi-scale knowledge graphs, enabling language models to integrate both broad context a…