Search papers, labs, and topics across Lattice.
3
0
3
6
DecompRL enables LLMs to solve complex problems by breaking them down into manageable sub-tasks, achieving a 50x reduction in GPU costs while enhancing solution diversity.
Extrapolating between code-generating RL agents trained on different unit test coverages unlocks better correctness-efficiency trade-offs than any single agent alone.
Reusing old data can actually *improve* LLM training, slashing compute costs without hurting performance.
