Search papers, labs, and topics across Lattice.
Amazon
1
0
3
LLM-generated survey responses can be statistically accurate yet still miss the option most preferred by humans, highlighting a critical flaw in current evaluation methods.