Qwen — Field Evidence
paperUnverified
Qwen · Qwen · complaint
Field note
Mergeable Model-Side Aggregation States for Long-Context Language Models — A known limitation of long-context language models is their increasingly unreliable performance in non-additive, set-based aggregation as context length grows. Examples include cardinality estimation, set
collected 2026-07-30original 2026-07-29
Does this shift the US–China race?
Be the first to call it
Impact on the index
Disputed benchmark · minor
China -15
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.