Llama — Field Evidence
paperUnverified
Llama · Llama · praise
Field note
GSM-Plus-BN: A Perturbation-Based Benchmark for Bangla Mathematical Reasoning in Large Language Models — The evaluation of mathematical reasoning in large language models (LLMs) has predominantly focused on high-resource languages like English. This has created a significant barr
collected 2026-07-16original 2026-07-14
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.