Fan Wu

DeepGuard introduces a novel multi-layer semantic aggregation framework to enhance secure code generation by collecting vulnerability cues from multiple upper layers of LLMs, significantly improving security while maintaining functional correctness.

The paper introduces FedRAG, a novel federated RAG framework that enables privacy-preserving cross-institutional knowledge collaboration by decoupling the self-attention mechanism from data localization using a specialized scrambling protocol.

This paper introduces a latent attack framework demonstrating that attacks can be embedded into the hidden representations of multi-agent systems, causing performance degradation even during clean, non-adversarial executions.

The paper introduces Cookie-Bench, a novel, autonomous, and reference-free evaluation framework that significantly improves the assessment of interactive web generation capabilities for frontier LLMs.

The paper proposes a novel zeroth-order optimization framework to enhance the robustness of LLM safety alignment, showing that few refinement steps can significantly improve safety while maintaining utility.

The paper introduces WebIGBench, a novel benchmark designed to rigorously evaluate multimodal LLMs' ability to generate code for complex, interactive webpages, addressing the limitations of existing static evaluation methods.

OmniOPD introduces a logit-free, chunk-level distillation framework that improves on standard On-Policy Distillation by using semantic similarity and peak-entropy scheduling, achieving state-of-the-art performance even with black-box teachers.

The paper proposes methods for detecting training instability in large language models using internal monitors based on the functional role of critical modules and earliest computational sites.

This paper proposes a method called selective prediction to enhance the reliability of large language models by allowing them to only predict for inputs where they are likely to be correct, reducing errors and enabling human-AI collaboration.

The paper introduces a memory agent to improve decision-making in long-horizon tasks by actively updating and intervening with reminders.

This paper proposes Gleam, a framework for efficient GPU sharing across local-area CUDA devices, reducing bandwidth overhead, improving API call latency, and ensuring context consistency.