Unverified50% confidenceFactExact time
Early interpretability research found that a certain neuron activated strongly to both cat faces and car fronts, and this phenomenon was called polysemanticity.
1
Sources
50%
Confidence
Long-term
Relevance
10/22/2026
First Seen
Sources
Was the Neural Network Always "Clean"? Interpreting Goodfire's Long-Form Explainability Essay
youtubeXiaoyu Li10/16/2026
Related Entities
Related Claims
UnverifiedOpenAI's 2020 circuit-mapping guide described this neuron as responding to "cat faces, car fronts, and cat legs".76% similarUnverifiedIn 2017, Google researchers Chris Olah, Alexander Mordvintsev, and Ludwig Schubert demonstrated a method: starting from random noise and repeatedly tweaking pixels to keep raising a certain neuron's activation value, the resulting images eventually contained cat faces, fox faces, and car fronts.69% similarUnverifiedNick Cammarata's analysis points out that InceptionV1's 4e:55 neuron actually detects a V-shaped geometric form, rather than cat faces or car fronts.67% similarUnverifiedAfter re-analysis, Camerota concluded that neuron 4E.55 is actually a V-shape detector, and that its "cat-car hybrid" appearance is an artifact artificially manufactured by the visualization optimization algorithm exploiting biases in the training data (many animal-face photos).66% similarUnverifiedIn image recognition, shallow neurons respond to low-level features such as edges and textures, while deep neurons respond to overall concepts.64% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1085037API
curl https://kongchang.com/api/v1/knowledge/claims/1085037MCP
get_claim(id=1085037)