Search papers, labs, and topics across Lattice.
Affiliation:
4
0
6
SymFold outperforms existing methods in protein inverse folding by effectively combining evolutionary and structural insights for sequence generation.
Unified multimodal models may excel in generation and understanding, but they often falter when reasoning about their own outputs, revealing hidden weaknesses in their capabilities.
Hard prompt compressors can leave critical context gaps, leading to a staggering 60% of examples suffering from referential dangling, which severely impacts accuracy in multi-hop question answering.
Ditch imperfect human annotations: this dual-reward RL approach trains image captioning models to be both more complete and more factually correct.
