Xiaohua Jia

发表论文 13 篇 · 总被引 73 次 · h-index 4

代表论文

• The Hidden Dimensions of LLM Alignment: A Multi-Dimensional Analysis of Orthogonal Safety Directions (2025 · International Conference on Machine Learning · 被引 33)
• Robot Collapse: Supply Chain Backdoor Attacks Against VLM-based Robotic Manipulation (2024 · 被引 19)
• The Hidden Dimensions of LLM Alignment: A Multi-Dimensional Safety Analysis (2025 · arXiv.org · 被引 18)
• M*: Every Task Deserves Its Own Memory Harness (2026 · arXiv.org · 被引 6)
• Can LLMs Refuse Questions They Do Not Know? Measuring Knowledge-Aware Refusal in Factual Tasks (2025 · arXiv.org · 被引 6)
• Towards Long-Horizon Interpretability: Efficient and Faithful Multi-Token Attribution for Reasoning LLMs (2026 · arXiv.org · 被引 3)