Search papers, labs, and topics across Lattice.
3
0
3
2
LAST achieves 95.4% accuracy with only 12.5% of visual tokens, revolutionizing edge-cloud inference efficiency.
Efficiently optimizing Gaussian capacity allocation can yield nearly 3 dB improvements in image representation quality without increasing the computational budget.
LUMI revolutionizes lossless image compression by decoupling it from tokenizer behavior, achieving superior performance with a unified approach across different LLM architectures.
