LatComm compresses LLM context once for reuse across models
On QuALITY the receiver sees 57 times less input and answers an article's questions three times faster.
4 stories, newest first · one card per story
On QuALITY the receiver sees 57 times less input and answers an article's questions three times faster.
RecCAR regularizes cross-attention so companion modalities better constrain joint video generation.
MCRL2 augments reinforcement learning with multi-resource cross-attention representations to improve cloud microservice scheduling and load balancing.
Researchers analyze the 'attention triangle' in audio-video diffusion models, showing bias-driven cross-attention routing causes semantic leakage and proposing inference-time interventions that improve grounding.
Today 1.6K readers · 2.4K views · 776 reads · all time 189K views