Search papers, labs, and topics across Lattice.
Institute of High Performance Computing, Agency for Science, Technology and Research, Singapore
2
0
4
Adjusting gate sharpness based on routing confidence leads to significant performance gains in Mixture-of-Experts models without the burden of extra parameters.
LLM safety filters can be bypassed by strategically fragmenting and camouflaging malicious intent across multiple turns, achieving a 26% improvement in jailbreak success rate on GPT-5-mini.
