Search papers, labs, and topics across Lattice.
5
0
6
0
Timing the entry of preference dimensions can lead to substantial performance gains in multi-preference alignment for LLMs.
Models trained with ACA-RL not only outperform on missing-premise tasks but also redefine how we evaluate reasoning under uncertainty in NLP.
T2I models struggle significantly with spatial instructions based on object orientation, achieving only 44.3% accuracy on frame-of-reference prompts.
Emotion-controllable portrait animation can now be achieved in real-time with a single training instance, revolutionizing the efficiency and expressiveness of animated avatars.
System prompts in commercial AI products are often a mixed bag, with 40% harboring instructions that can undermine user interests, revealing a critical gap in accountability.
