Search papers, labs, and topics across Lattice.
4
1
7
2
Self-looping reasoning in LLMs can be effectively mitigated through targeted hidden-state interventions, leading to improved reasoning quality and efficiency.
Training data diversity is the secret sauce that boosts agentic model performance, with OpenThoughts-Agent achieving a notable accuracy leap over existing benchmarks.
Rollout design in LLM reinforcement learning is more than just sampling trajectories – it's a modular pipeline you can optimize for reliability, coverage, and cost.
Existing music editing systems often compromise on preserving key musical elements, but MuseCPEval provides a robust framework to ensure these facets remain intact during editing.
