Unverified50% confidenceFactExact time
In the Inception V1 model, the uses of about 0-5% of the neurons in the last layer can be understood, which is one of the main pieces of evidence for superposition theory.
1
Sources
50%
Confidence
Long-term
Relevance
10/17/2026
First Seen
Sources
Neural Networks Were Always Clean? Goodfire Researcher Overturns Superposition Theory
bilibili迦太基盐商10/16/2026
Related Entities
Related Claims
UnverifiedTraditional models of nervous system evolution tend to depict a single lineage: from simple nerve nets to centralized ganglia, and then to a highly organized central nervous system.67% similarUnverified贝叶斯定理在朴素贝叶斯分类器、贝叶斯神经网络、变分自编码器（VAE）和扩散模型中都有直接应用66% similarVerifiedMechanistic interpretability attempts to understand what concepts each neuron and each layer's connections encode inside a neural network.64% similarVerifiedThe o1 model performs internal chain-of-thought reasoning before answering, far surpassing GPT-4o on tasks such as mathematics, programming, and scientific reasoning.64% similarVerifiedInterpretability research from MIT and Anthropic found that there are large numbers of polysemantic features inside models, with a single neuron potentially responding simultaneously to multiple seemingly unrelated semantic concepts.63% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1052463API
curl https://kongchang.com/api/v1/knowledge/claims/1052463MCP
get_claim(id=1052463)