Search papers, labs, and topics across Lattice.
1
0
2
17
Open-ended generation models are failing to capture factual completeness, with top performers only achieving 58.7% accuracy on a new benchmark designed to assess this critical aspect.