Search papers, labs, and topics across Lattice.
Jilin University
6
0
9
Hydrogen's 1s electron can drive ferromagnetism, challenging long-held beliefs about magnetic materials.
Training LLMs with a compact rolling memory can lead to more robust reasoning, outperforming models that rely on full historical context.
Language complexity is not just about length; the ladderpath index reveals deep structural invariances across diverse languages.
Achieving unsupervised day-night re-identification with accuracy rivaling fully supervised methods, this framework eliminates the need for costly annotations.
Reinforcement learning can significantly enhance adaptive sampling in large language models, leading to better performance with fewer resources.
Current MLLMs falter in multi-stream contexts, achieving only 50% effectiveness, revealing a critical gap in their design for real-world applications.
