Search papers, labs, and topics across Lattice.
New York University
1
0
1
LLM benchmark scores are far more sensitive to prompt phrasing, judge model, and temperature than previously thought, creating exploitable surfaces and undermining confidence in published results.