Unverified50% confidenceBenchmarkExact time
On the task of distinguishing AI text from human text, both detector types reached a maximum text-level macro-average F1 of 0.96 and a maximum sentence-level score of 0.95, and were unaffected by spelling errors.
1
Sources
50%
Confidence
Long-term
Relevance
11/10/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedThe decision thresholds of AI-generated text detectors are usually set on long texts, and applying them directly to short texts causes the false positive rate to drift.71% similarUnverifiedThe text generated by modern LLMs is increasingly difficult to distinguish from human writing in terms of grammar and style, posing accuracy challenges for AI detection tools based on textual features (such as GPTZero and ZeroGPT).69% similarUnverifiedCombining perplexity, burstiness, and word-frequency anomaly analysis significantly improves AI text detection accuracy compared to any single method.68% similarVerifiedAfter text is manually rewritten or polished, the accuracy of AI detection methods based on textual features drops significantly.68% similarUnverifiedAI text detectors identify differences between human writing and machine-generated content from the statistical features, linguistic patterns, and semantic structures of text.67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1154835API
curl https://kongchang.com/api/v1/knowledge/claims/1154835MCP
get_claim(id=1154835)