Search papers, labs, and topics across Lattice.
College of Computing and Data Science, Nanyang Technological University
2
0
4
Neglecting edge information in graph-aware LLMs leads to significant performance bottlenecks, but the CureLLM framework unlocks superior alignment by leveraging curvature-enhanced graph representations.
Forget static prompts and rigid workflows: OrchMAS dynamically orchestrates heterogeneous LLM agents for scientific reasoning, achieving superior performance through iterative replanning and role reallocation.
