Search papers, labs, and topics across Lattice.
1
0
2
Revised commonsense benchmarks don't boost predictive power for downstream tasks, revealing a critical limitation in their utility for assessing LLM capabilities.
