Search papers, labs, and topics across Lattice.
5
0
6
7
Mechanist uncovers a surprising safety risk where unsafe traits can transfer across modalities, challenging assumptions about training data safety.
Unsupervised skill discovery can boost data-analytic agent performance by over 30% without the need for labeled data.
SkillAdaptor achieves targeted skill updates that enhance LLM performance by over 1.5 points on key benchmarks, revolutionizing how agents adapt to failures in real-time.
Forget manual data curation – now LLMs can autonomously engineer training data that boosts student model performance by over 57%.
LLMs often fail to maintain accurate beliefs in multi-turn interactions, but targeted reinforcement learning and representation steering can dramatically improve their contextual reasoning.
