Search papers, labs, and topics across Lattice.
Affiliation:
2
0
4
Instruction tuning degrades the latent moral geometry of LLMs, even though distribution-driven steering vectors can naturally mirror human value topologies to enable coherent cross-value generalization.
State-of-the-art Vision-Language Models may excel in understanding but falter dramatically in creative tasks and factual recall, revealing hidden cognitive weaknesses.
