Search papers, labs, and topics across Lattice.
4
0
6
4
SkillCAT boosts LLM agent performance by over 40% through a novel approach that intelligently evolves skills without retraining the model.
Multi-hop question answering gets a serious boost: ConRAG's consensus-driven, multi-view retrieval approach leapfrogs existing RAG methods by a significant margin.
Video-LLMs are leaving performance on the table: explicitly anchoring to keyframes before answering questions unlocks significant gains in Video TextVQA.
LLMs struggle to retrieve the right tools when instructions are vague, but a simple "bridge model" that rephrases instructions can more than double retrieval accuracy.
