Zhen Huang

This paper proposes using transformer-based models on program slices to accurately detect C/C++ software vulnerabilities by capturing both local and global contextual information.

The paper proposes a novel attack paradigm demonstrating how compromising a single robot in an LLM-controlled multi-robot system can rapidly propagate malicious intent to cause coordinated unsafe actions across the entire system.

The paper proposes Skill-Conditioned Gated Self-Distillation (SGSD), a novel framework that uses retrieved, potentially noisy skills to guide LLM reasoning, achieving state-of-the-art performance on mathematical reasoning benchmarks.

LoopFM proposes a novel framework to significantly improve knowledge distillation for recommendation systems by structuring the rich intermediate embeddings of large foundation models as input features, thereby overcoming the limitations of single-scalar prediction transfer.

The paper introduces RHELM, a new benchmark designed to test LLMs' long-term memory by simulating realistic, complex, and evolving dialogues that integrate multiple heterogeneous data sources.

SkillSmith is a synergy-aware framework that jointly co-evolves skills and tools, significantly improving self-improving agent systems by modeling skill-tool interactions and diagnosing failures.

AdaCodec introduces a predictive visual coding scheme for video MLLMs, significantly improving efficiency and performance by transmitting only inter-frame changes and full reference frames when necessary.

The paper proposes DIST-FL, a distributed system using multiple TEEs and an append-only ledger to enhance the security and robustness of federated learning aggregation against server-side adversaries.