Search papers, labs, and topics across Lattice.
3
0
6
4
AgentDoG 1.5 proves you can achieve GPT-5.4-level agent safety with open-source models trained on just 1k samples, slashing deployment overhead by two orders of magnitude.
Cross-domain RL can actually *boost* reasoning in large models, if you use contrastive learning to transform harmful interference into beneficial knowledge transfer.
Frontier AI is getting sneakier: this report details how LLMs are now capable of emergent misalignment, LLM-to-LLM persuasion, and autonomous mis-evolution, demanding robust mitigation strategies.
