Search papers, labs, and topics across Lattice.
Beijing Institute of AI Safety and Governance, China., Beijing Key Laboratory of Safe AI and Superalignment, China., BrainCog Lab, Institute of Automation, Chinese Academy of Sciences, China.
4
0
10
Optimize prompts for multi-agent systems using LLMs with a new bandit-based approach that uses GNNs to capture topology-aware prompt semantics, achieving state-of-the-art performance.
Unlock the secrets hidden in complex X-ray diffraction data with WPEM, a new AI-driven refinement workflow that outperforms existing methods by explicitly enforcing Bragg's law.
Learning-based spatial reasoning gets a boost by using geometry not just for refinement, but as a critical arbiter that validates and absorbs learned geometric observations.
Frontier AI models exhibit widespread safety vulnerabilities across multiple pillars, including risky agentic autonomy and catastrophic risks, according to a new comprehensive benchmark.
