Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
2
0
6
Context compaction in agentic RL can run up to 5x faster simply by streaming the KV cache instead of flushing it—accidentally turning standard LLMs into recurrent agents that preserve evicted context purely through RL.
TREAD transforms existing robot learning datasets into powerful tools for generalization, enabling robots to tackle unseen tasks with improved instruction-following capabilities.