Unverified50% confidenceFactExact time
ARC-AGI基准由Keras创始人François Chollet于2019年提出，每道题都是独一无二的视觉图案变换任务，用于衡量抽象推理能力而非模式匹配
1
Sources
50%
Confidence
Long-term
Relevance
8/25/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedThe design philosophy of plug-and-play modules can be traced back to the concentrated exploration of attention mechanisms in the computer vision field between 2017 and 2018.64% similarUnverifiedThis workflow seamlessly connects human visual annotation capabilities with AI image generation capabilities, resolving the ambiguity problems that arise when describing modification requirements in plain text63% similarUnverifiedARC-AGI-V2 increases combinatorial complexity and visual-spatial reasoning difficulty compared to the previous version.62% similarUnverified参考图的引入激活了模型的图像到图像推理能力，将真实建筑视觉特征与叙事需求对齐，减少AI感违和感61% similarUnverifiedK3 was released alongside the Perception Bench benchmark, which uses reverse design to distill 10 atomic visual perception capabilities from more than 40 existing benchmarks, comprising 3,000 manually verified high-difficulty questions.60% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/798190API
curl https://kongchang.com/api/v1/knowledge/claims/798190MCP
get_claim(id=798190)