Search papers, labs, and topics across Lattice.
School of Computing and Artificial Intelligence, Southwest Jiaotong University
3
0
7
Forget same-family constraints: you can compress prompts for LLaMA with a Qwen draft model and still get 90-100% of the original performance.
Ditch pixel-by-pixel processing: C$^2$SSM achieves state-of-the-art UHD image restoration by scanning semantic clusters, not individual pixels, for a massive efficiency boost.
LLMs can serve as effective auxiliary teachers in knowledge distillation, significantly boosting the performance of lightweight temporal knowledge graph reasoning models.
