Search papers, labs, and topics across Lattice.
1
0
3
Forget contrastive learning: LLM2Vec-Gen trains special tokens to represent the *response* an LLM would give, yielding SOTA embeddings with better safety and reasoning.
