Bo Zhang

The paper introduces TrojanMerge, a framework demonstrating that model merging can be exploited to systematically compromise the safety alignment of multiple individually safe LLMs.

AttnDiff introduces a data-efficient white-box framework that extracts intrinsic attention-based fingerprints to verify the provenance and detect unauthorized derivation of large language models (LLMs) despite common model laundering techniques.

The paper proposes an operation-centric, TEE-backed isolation model to constrain self-hosted computer-use agents, preventing malicious or unsafe host-level operations without sacrificing general functionality.

MemMark introduces a state-evolution attribution watermark that embeds owner-controlled signals into latent memory-write decisions, enabling robust provenance tracking for agent memory even when all traditional logs and metadata are lost.

The paper introduces a histogram-regularized latent diffusion model to synthesize highly realistic and subtype-specific pulmonary nodules in 3D CT volumes, addressing the limitations of existing methods that fail to capture accurate lesion-level intensity distributions.

The paper introduces AgentSchool, an advanced LLM-powered multi-agent simulator that models learning as state transitions to provide a robust, ethically viable testbed for educational research and pedagogical reform.

The paper introduces AgentDoG 1.5, a lightweight and scalable alignment framework that significantly improves AI agent safety and security for complex, open-world agentic scenarios.

The paper introduces AgentDoG 1.5, a lightweight and scalable alignment framework that significantly improves AI agent safety and security for complex open-world agent deployments.

DataShield proposes an efficient method to identify safety-degrading samples within benign datasets, quantifying each sample's contribution to an LLM's compliance behavior.

The paper introduces AMix-2, a novel protein-text foundation model that unifies protein understanding and sequence design by embedding both modalities in a shared token space, achieving state-of-the-art performance on comprehensive benchmarks.

The paper proposes DiReCT, a novel framework that treats data selection during LLM annealing as a constrained optimization problem based on the spectral geometry of the loss landscape, achieving state-of-the-art performance.

DataShield proposes an efficient method to identify safety-degrading samples within benign datasets, preventing the degradation of LLM safety capabilities during fine-tuning.

SeClaw is a new framework that synthesizes security tasks from structured risk specifications to evaluate autonomous LLM agents' behavior in stateful environments, focusing on the process of unsafe actions rather than just the final outcome.

SeClaw is a new framework that uses specification-driven task synthesis to create comprehensive and controllable security benchmarks for evaluating the unsafe behaviors of autonomous LLM agents.

MLEvolve is a novel self-evolving multi-agent framework that enables LLM agents to discover and optimize machine learning algorithms for complex, long-horizon tasks.

This paper introduces Agents-K1, an end-to-end knowledge orchestration pipeline that converts raw documents into agent-native scientific knowledge graphs.