Search papers, labs, and topics across Lattice.
3
0
5
1
Stabilizing RL training is now possible by modulating importance ratios with hazard-aware penalties, preventing both mode collapse and policy erosion.
Achieve a remarkable 12.4x speedup in 3D reconstruction by mimicking the efficiency of keypoint matching with a novel dual-branch attention mechanism.
Forget expensive, full-parameter fine-tuning: surgically correcting just 4k errors in LLM reasoning boosts accuracy by 6.2% while preserving prior knowledge.
