Unverified50% confidenceOpinionExact time
Simply scaling up models or improving general coding ability may not be able to break through the last-mile problem of long-document extraction.
1
Sources
50%
Confidence
Long-term
Relevance
10/14/2026
First Seen
Sources
Related Entities
Related Claims
Unverified简单堆砌上下文长度并不能线性提升长任务性能，因存在中间丢失问题74% similarUnverifiedResidual amortization error under rare settings cannot be eliminated by simply scaling up the model.74% similarUnverifiedSwitching to a better metric cannot salvage an analysis that already contains arbitrariness at the data construction stage.72% similarUnverifiedA single throughput figure (tok/s) cannot reflect a model's true capability, long-context cost, or cross-model comparability.72% similarUnverified代理化场景对模型的工具调用和长程规划能力要求更高，而这些维度是传统代码生成基准难以充分衡量的72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1023625API
curl https://kongchang.com/api/v1/knowledge/claims/1023625MCP
get_claim(id=1023625)