Containment Verification: Ensuring AI Safety Without Alignment
Discover containment verification, a new AI safety method guaranteeing secure AI operations independent of model alignment or external conditions.
UxSID: Semantic User Interest Modeling for Ultra-Long Sequences
Discover UxSID, a semantic-aware framework improving user interest modeling for ultra-long sequences, boosting ad performance and user engagement.
SearchSkill: Boost LLM Search with Evolving Skill Banks
Discover how SearchSkill improves LLMs' query generation using dynamic skill banks for accurate, efficient search and better question answering.
CATO: Efficient Neural PDE Solver with Charted Attention
Discover CATO, a novel neural operator using charted attention for efficient, accurate PDE solutions on complex geometries with reduced computational costs...
Re$^2$Math: Benchmarking Theorem Retrieval in Math Research
Explore Re$^2$Math, a benchmark evaluating theorem retrieval in research-level mathematics using AI and language models for improved proof assistance.