Search papers, labs, and topics across Lattice.
4
0
5
2
Game solvers can teach LLMs how to make better decisions in long-horizon tasks by providing actionable turn-level feedback, leading to superior performance in complex environments.
Agent-UCT slashes logical search costs by over 73% while optimizing agentic workflows, making it a game-changer for RAG pipelines.
Dense prediction rewards can backfire catastrophically in GRPO-trained LLMs, leading to perfect predictions but zero task success.
Agents that ace long-context recall can still bomb when they need to use that memory to actually *do* something, revealing a critical flaw in how we currently evaluate memory in AI.
