Search papers, labs, and topics across Lattice.
4
0
7
0
RoboDojo reveals that existing benchmarks fail to capture the full spectrum of robot manipulation capabilities, paving the way for more robust evaluations that bridge the gap between simulation and real-world performance.
Compressing multimodal LLM inputs with learned memory-augmented distillation not only speeds up inference by 1.5x, but also *improves* accuracy on audio-visual question answering.
OmniRetriever-7B leapfrogs Gemini Embedding 2 in zero-shot audio-video-text retrieval, proving that smart training objectives can close the gap with closed-source giants.
Current image quality metrics struggle to articulate *why* one high-quality image is better than another, but this challenge shows MLLMs are closing the gap by providing expert-level explanations.
