Unverified50% confidenceFactExact time
The key metric, the headline score, refers to the proportion of sentences fully and correctly restored; if even one tone mark is wrong the entire sentence is counted as incorrect, and the test is built on the kaggle-benchmarks library.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/11/2026
First Seen
Valid until: 1/9/2027
Sources
Related Entities
Related Claims
UnverifiedThe benchmark is built on the Kaggle Benchmarks platform; each model is required to answer 287 scored questions spanning five dimensions: numbers, verbs, translation, cloze, and error correction.71% similarUnverifiedPrompt caching is based on prefix matching, reusing already-computed KV Cache to skip redundant computation, so fixed content at the beginning of prompts should be kept consistent to maximize cache hit rate.61% similarUnverified使用思维链（CoT）提示要求裁判先分析再打分可将裁判与人工标注的一致性提升约7-12%61% similarUnverifiedA benchmarking mechanism should be established that uses standard questions with known answers to continuously verify the accuracy of Agent outputs.61% similarUnverified锚定样本技术在评估提示词中加入已由人类专家标注明确分数的示例，覆盖评分量表各个档位，可帮助裁判模型准确理解评分标准61% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/999187API
curl https://kongchang.com/api/v1/knowledge/claims/999187MCP
get_claim(id=999187)