Search papers, labs, and topics across Lattice.
8
0
12
0
Current pathology VLMs can achieve high accuracy without visual input, revealing critical flaws in how we assess their multimodal capabilities.
Compact models like Athena-Brain-8B can outperform larger counterparts in embodied tasks while maintaining strong general intelligence.
Humanoid controllers can achieve better performance on challenging motions with a compact pipeline of capability-aligned policy experts, reducing the need for extensive training data.
Kullback鈥揕eibler divergence is the secret sauce that ensures density-level and sample-conditioned objectives align perfectly in generative modeling.
Roken can generate coordinated multi-robot trajectories in real-time, achieving higher success rates than traditional sequential planning methods.
Surprisingly, the "think before answer" paradigm fails to enhance generative recommendation models, prompting a novel approach that redefines how reasoning is integrated into these systems.
Deep learning models can now classify complex nuclear reaction events with high accuracy, even correcting errors made by traditional methods.
Aggregate benchmark scores can be misleading: models with statistically indistinguishable atomic knowledge can exhibit composition behavior differences exceeding 40 percentage points.
