GPT — Field Evidence
paperUnverified
GPT · GPT · praise
Field note
Position Bias is Hidden Behind Ceiling Effects: A Permutation Diagnostic for LLM Benchmarks — Position bias in multiple-choice LLM evaluation is widely cited as a confound in capability comparisons, but published measurements rely on single answer-order shuffles whose results con
collected 2026-07-24original 2026-07-23
Does this shift the US–China race?
Be the first to call it
Impact on the index
Benchmark win · minor
US +30China -7
Directional contribution — before recency decay and per-type diminishing returns. How it works →
Related Front
US vs ChinaLikely
U.S. FrontiervsChina Open-Weight
Likely — capability gap narrowing on common tasks
Letters from the Front
Be the first to file a report.