Search papers, labs, and topics across Lattice.
6
0
9
2
AI reviewers can be gamed by merely altering how research is presented, achieving significant score increases without changing the underlying science.
DB-Nav redefines object navigation by leveraging relational biases to filter out unreliable semantic cues, achieving superior performance without the overhead of complex vision-language reasoning.
Achieving nearly one-pixel precision in lunar rover localization without GNSS signals could revolutionize autonomous exploration on the Moon.
Current video MLLMs struggle to grasp fleeting visual events, with top models barely surpassing 39% accuracy on critical momentary tasks.
Stop rewarding reasoning that just looks good – reward reasoning that actually *helps* the downstream model solve the task.
Image editing benchmarks are broken: even GPT-4 is worse than the new PVC-Judge model at assessing visual consistency in edited images.
