Search papers, labs, and topics across Lattice.
7
0
10
0
No memory substrate is a one-size-fits-all solution; the right choice depends on the task, with broad retrieval boosting QA but hindering decision-making.
Static token credit misaligns with evolving training dynamics, but Se-DPO's adaptive approach boosts performance by nearly 10 points on key benchmarks.
Adaptive memory management can boost LLM task success by over 15 points while slashing token usage by up to 20%.
Robot actions can serve as powerful geometric supervision, enabling sparse 3D representations that are both reusable and effective across diverse manipulation tasks.
MT-EditFlow bridges the gap between local planning and global success in multi-turn image editing, achieving a significant performance boost over leading models.
Counterintuitively, distilling LLMs is more effective when you only use the first few tokens of a student's rollout, surpassing full-trajectory distillation while saving compute.
Forget hand-designed agent communication topologies: Agent Q-Mix learns decentralized communication strategies that boost accuracy and token efficiency in LLM multi-agent systems.
