Search papers, labs, and topics across Lattice.
University of Chinese Academy of Sciences
3
0
6
AlphaToken reveals a method to enhance LLM post-training by strategically valuing response tokens, leading to improved performance and reduced forgetting.
Forget cloud-based LLM hallucinations: G-DRAGON grounds outdoor robot navigation in lightweight, locally-retrieved OSM data for robust long-range planning and precise last-mile exploration.
FOCAL-Attention resolves the inherent coverage-anchoring conflict in heterogeneous graph learning, outperforming existing methods in multi-label node classification.
