Unverified50% confidenceFactExact time
In 2022, Anthropic verified with small-scale networks that superposition does exist, but for large real networks the evidence is only "consistent with superposition".
1
Sources
50%
Confidence
Long-term
Relevance
10/22/2026
First Seen
Sources
One Neuron Detects Both Cats and Cars? We Reproduced It and Found a New Explanation
youtubeUp the Tech Tree10/16/2026
Related Entities
Related Claims
UnverifiedThe universal approximation theorem for neural networks states that a sufficiently wide or sufficiently deep network can approximate any continuous function, but does not answer the question of what network size is required to achieve a specific accuracy.67% similarUnverifiedThe paper proves that, under appropriate network depth and width configurations, the convergence rate of dense ReLU networks can overcome the curse of dimensionality, achieving polynomial-level sample complexity.63% similarUnverifiedThe lottery ticket hypothesis was proposed by Frankle and Carlin in 2019, pointing out that within a randomly initialized large neural network there exists a smaller subnetwork that, when trained in isolation, can achieve performance comparable to or even better than the original network.62% similarUnverifiedThe experiments used an idealized Provenance Density interface; large-scale claim-level evidence verification and consistency checking in the real world remain enormous engineering challenges.62% similarUnverifiedEven with batch normalization already included in the plain-56 network, severe degradation still occurred, indicating that batch normalization alone cannot solve the optimization difficulties of deep networks, and the contribution of skip connections is independent and irreplaceable.62% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1085051API
curl https://kongchang.com/api/v1/knowledge/claims/1085051MCP
get_claim(id=1085051)