Unverified50% confidenceBenchmarkExact time
In the experiments, the classical weight-space bound rapidly inflated as the number of parameters increased, exceeding 940% on the largest model and completely failing.
1
Sources
50%
Confidence
Long-term
Relevance
10/16/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedThe computational complexity of diffusion models in the temporal dimension grows approximately quadratically with duration, and memory and compute consumption rise sharply.63% similarUnverifiedWhen the hidden dimension reaches 1024, attention logits occasionally spike dramatically, becoming a stability risk after scaling up.63% similarUnverifiedAs large model reasoning capabilities strengthen, the tokens consumed per single call rise significantly, creating a paradox of 'stronger model' versus 'faster limit reached' under fixed usage limits.62% similarUnverifiedWhen model scale exceeds a certain threshold, certain capabilities suddenly jump from near zero to a significant level, rather than growing linearly with parameter count.62% similarUnverifiedA deviation of one ULP (unit in the last place), amplified through tens of thousands of steps of backpropagation, is sufficient to send the model into a completely different parameter space.61% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1046147API
curl https://kongchang.com/api/v1/knowledge/claims/1046147MCP
get_claim(id=1046147)