Search papers, labs, and topics across Lattice.
5
0
5
Most enrichment strategies in RAG systems backfire, with a smaller model outperforming a larger one by 19 F1 points when metadata aligns with model capabilities.
Local-first information retrieval can achieve cloud-level answer quality while keeping sensitive data on user devices, redefining the tradeoff between search scope and quality.
Stop retrieving passages in your RAG system: NuggetIndex shows that retrieving and filtering atomic "nuggets" of information yields substantial gains in recall, temporal correctness, and reduced conflicts.
Forget synthetic QA datasets – AgentSim offers verifiable, step-by-step RAG traces, revealing how LLMs *actually* reason over documents.
The core assumption that user behavior reveals intent breaks down when AI agents are privately configured by humans, creating a fundamental identifiability problem for information retrieval.
