Search papers, labs, and topics across Lattice.
2
0
2
Systematic gaps in AI evaluation reporting are exposed, revealing inconsistencies that hinder reliable comparisons across thousands of models and benchmarks.
Leaderboard rankings are more noise than signal: contributor metadata matters more than architecture, and scaling laws are unreliable.
