Search papers, labs, and topics across Lattice.
5
0
10
0
Autonomous driving gets a boost: EvoDriveVLA's collaborative perception-planning distillation framework significantly enhances VLA model performance by tackling perception degradation and planning instability.
Freezing your feature extractor during online category discovery leads to suboptimal performance: TALON shows how test-time adaptation can substantially boost novel class accuracy and prevent category explosion.
LLMs excel at compliance-focused grant review but struggle with holistic assessment, revealing a critical gap in their ability to replace human expert panels.
Mobile agents trained with RL struggle to generalize to new app interfaces, improving only 8.3% compared to supervised baselines, despite a 26.1% gain on unseen task instances.
By strategically warming up residual connections layer-by-layer, ProRes unlocks faster and more stable pretraining for language models.
