Unverified50% confidenceFactExact time
Anthropic、DeepMind等机构已将SAE成功应用于GPT类语言模型，发现了大量具有可解释语义的特征方向
1
Sources
50%
Confidence
Long-term
Relevance
7/9/2026
First Seen
Sources
Block-Sparse Featurizers: Opening the Black Box of Vision Model Neural Geometry
hackernewshackernews7/7/2026
Related Claims
UnverifiedMainstream large models such as the GPT series, LLaMA, and DeepSeek are all built on the autoregressive language model paradigm.76% similarVerifiedThe underlying architecture of GPT-Image-2 integrates the semantic understanding capabilities of large language models, enabling it to automatically convert natural language descriptions into visual generation instructions.76% similarUnverifiedThe GPT series follows the decoder route, adhering to the autoregressive language model objective, and has revealed the phenomenon of emergent capabilities through scaling.75% similarVerifiedLarge language models such as GPT and BERT are based on the self-supervised learning paradigm75% similarVerifiedCurrently mainstream large language models include GPT-4, Claude, DeepSeek, Tongyi Qianwen, and others.75% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/313865API
curl https://kongchang.com/api/v1/knowledge/claims/313865MCP
get_claim(id=313865)