14 papers
History Replays Itself: Unleashing Long-Context Capability in Linear Attention
Li Nanbo, Firas Laakom, Yilan Zhang +10
Linear attention promises the efficiency needed for long-context language modelling, yet its fixed-size recurrent state must simultaneously support ongoing computation and preserve…
An unbounded convex set with the fixed point property for nonexpansive mappings
M. Japón, Javier Laguna-Ricart
We construct an equivalent norm on and a nonempty closed convex unbounded subset having the fixed point property for nonexpansive self-mappings. The ambien…
Mitigating LLM Over-Refusal via Dynamic Semantic Routing Calibration
Zixuan Wang, Bingjie Zhang, He Zhao +1
Large language models (LLMs) aligned for safety often suffer from over-refusal, incorrectly rejecting benign yet safety-related instructions. Prior studies primarily attribute this…
Bayesian Sparse Low-Rank Adaptation for Large Language Model Uncertainty Estimation
Jijie Zhang, Zhe Ren, Quan Zhang +1
Large language models (LLMs) exhibit remarkable reasoning capabilities, but their task-specific fine-tuning is notoriously plagued by overconfidence, severely hindering trustworthy…
SpecGradFilter: A Spectral Gradient Filtering Framework for Taming Federated Heterogeneity
Liyang Yuan, Yibo Yang, Dandan Guo +2
Federated Learning (FL) is fundamentally challenged by statistical heterogeneity, where non-identically distributed (non-IID) data induces client drift that severely hampers global…
Self-Improvements in Modern Agentic Systems: A Survey
Zhe Ren, Yimeng Chen, Dandan Guo +9
Self-improving autonomous agents are moving from research prototypes to deployed systems. The primary goal is controllable evolution, or adaptation, from experience with minimal or…
FedFFT: Taming Client Drift in Federated SAM via Spectral Perturbation Filtering
Liyang Yuan, Yibo Yang, Dandan Guo
Federated Learning (FL) enables decentralized training without data sharing, but suffers from statistical heterogeneity across clients, leading to client drift, poor generalization…
Defending Against Harmful Supervision Hidden in Benign Samples
Bang An, Yibo Yang, Dandan Guo +3
Existing defenses are effective when harmful content is explicitly mixed into downstream fine-tuning data, but crafted samples can instead hide harmful supervision inside benign ta…
How Much Can We Trust LLM Search Agents? Measuring Endorsement Vulnerability to Web Content Manipulation
Yimeng Chen, Zhe Ren, Firas Laakom +3
Large language model (LLM)-based search agents synthesize open-web content into actionable recommendations on behalf of users, creating a risk that attacker-published pages are tra…
GateMem: Benchmarking Memory Governance in Multi-Principal Shared-Memory Agents
Zhe Ren, Yibo Yang, Yimeng Chen +7
Memory benchmarks for LLM agents largely assume single-user settings, leaving shared assistants for hospitals, workplaces, campuses, and households understudied. In these deploymen…
H2HMem: A Multimodal Memory Benchmark for Agents in Human-Human Interactions
Shiping Zhu, Yibo Yang, Zhengyang Wang +3
Large language model agents are increasingly deployed in human-human interaction settings, such as meeting assistants and clinical documentation systems, where they must observe co…
Safeguarding LLM Fine-tuning via Push-Pull Distributional Alignment
Haozhong Wang, Zhuo Li, Yibo Yang +3
The inherent safety alignment of Large Language Models (LLMs) is prone to erosion during fine-tuning, even when using seemingly innocuous datasets. While existing defenses attempt…
PDR: A Plug-and-Play Positional Decay Framework for LLM Pre-training Data Detection
Jinhan Liu, Yibo Yang, Ruiying Lu +4
Detecting pre-training data in Large Language Models (LLMs) is crucial for auditing data privacy and copyright compliance, yet it remains challenging in black-box, zero-shot settin…
A Guardrail for Safety Preservation: When Safety-Sensitive Subspace Meets Harmful-Resistant Null-Space
Bingjie Zhang, Yibo Yang, Zhe Ren +4
Large language models (LLMs) have achieved remarkable success in diverse tasks, yet their safety alignment remains fragile during adaptation. Even when fine-tuning on benign data o…