Expired70% confidenceFactExact time
Researchers have identified statistical anomalies indicative of benchmark contamination in mainstream models such as GPT-4 and Claude.
2
Sources
70%
Confidence
Medium-term (~90 days)
Relevance
7/5/2026
First Seen
Valid until: 10/3/2026(expired)
Sources
Related Claims
Unverified2023年多项学术研究证明GPT-4、Claude等主流大模型均对提示注入攻击存在不同程度的脆弱性75% similarUnverifiedResearch from institutions including ETH Zurich and Google DeepMind has demonstrated that top-tier models such as GPT-4 and Claude exhibit significant vulnerability to adversarial prompts, with success rates sometimes exceeding 50%.68% similarUnverifiedDifferent large models (GPT-4, Claude, Llama, etc.) differ in their response characteristics to prompts, and in practice it is necessary to fine-tune for specific models.68% similarUnverifiedBenchmark data released after the launch of GPT-5 show that it nearly comprehensively surpasses Claude and all other models.66% similarUnverified随着GPT-4、Claude等模型能力趋于收敛，单纯依靠模型规模提升带来的边际收益已在递减66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/112489API
curl https://kongchang.com/api/v1/knowledge/claims/112489MCP
get_claim(id=112489)