Unverified60% confidenceFactExact time
Kosinski's (2023) research shows that GPT-4 performs close to human level on standard false-belief tasks.
2
Sources
60%
Confidence
Long-term
Relevance
7/10/2026
First Seen
Sources
Forget Benchmark Scores! Four Practical Standards to Judge an LLM's Real Capability
bilibili鲲鹏AI探索局7/6/2026
Related Claims
UnverifiedMainstream large language models including GPT-4, Gemini, and Grok have all had hallucination behavior documented by researchers74% similarUnverifiedResearchers including Tal Ullman at New York University found that GPT-4's performance dropped sharply when classic theory-of-mind tests were modified slightly, suggesting the model may be performing surface-level pattern matching.73% similarUnverifiedLarge models such as GPT-4 have demonstrated some vulnerability analysis capability on CTF challenges, but their accuracy in real-world complex codebases remains far below that of human experts.72% similarUnverified2023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70% similarUnverifiedEven GPT-4-level models cannot fundamentally eliminate hallucinations.70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/426483API
curl https://kongchang.com/api/v1/knowledge/claims/426483MCP
get_claim(id=426483)