Wei Lu

The paper proposes GeoMark, a geometry-aware localized watermarking framework that robustly protects Embedding-as-a-Service (EaaS) against model stealing and copyright infringement while preserving utility.

The paper introduces Canonical-Context On-Policy Distillation (CCOPD) to improve multi-turn language model performance by mitigating 'self-anchored drift,' ensuring consistent answers regardless of whether the evidence is presented in a single prompt or gradually across multiple turns.

This paper introduces 'implicit identity' as a unifying framework to survey and categorize LLM fingerprinting and watermarking techniques for verifying ownership and provenance across datasets, models, and generated content.

The paper introduces DiscourseFlip, a novel graph-guided attack that demonstrates how coordinated poisoning across a multi-topic query space can manipulate the overall opinion generated by black-box Retrieval-Augmented Generation (RAG) systems.

The paper introduces DiscourseFlip, a novel black-box, graph-guided attack that manipulates opinions across an entire multi-topic query network, demonstrating a significant leap in scope and effectiveness over existing RAG attack methods.

The paper introduces Grounded Agentic Interaction Synthesis (GAIS), a framework that generates high-quality, diverse, and complex agentic training data by anchoring tasks to real-world protocols, significantly improving base model performance.

The paper introduces State-Adaptive Prompt Optimization (SAPO), a novel training strategy that treats prompts as dynamic variables to achieve robust fine-tuning, significantly mitigating catastrophic forgetting and improving generalization in LLMs.

The paper introduces the threat model of sequential data poisoning, demonstrating that multiple, collaborating attackers can exploit compound vulnerabilities in LLM post-training pipelines that are invisible when analyzing individual stages.