Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
Affiliation:
1
0
2
0
Standard accuracy metrics can inflate language model performance by up to 0.159 points, drastically misrepresenting their true competence in convex grading systems.