Unverified50% confidenceFactExact time
Adversarial hardening narrows the class gap between NPC and P twins from 24.8 percentage points to 14.6 percentage points, suggesting that part of the surface complexity gap stems from models' familiarity with textbook wording.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/22/2026
First Seen
Valid until: 1/20/2027
Sources
Related Entities
Related Claims
UnverifiedWith the number of categories held constant, changing which categories are compared shifts the distance between two representations by more than twice the measured effect.68% similarUnverifiedThe H₁(P) complexity measure is a weighted sum of the reciprocal squared gaps between each suboptimal arm and the optimal arm.65% similarUnverifiedLow-rank decomposition and quantization compression are orthogonally complementary dimensions; quantizing the smaller matrices A and B separately results in smaller errors, and the combined compression ratio can reach several times that of either method alone64% similarUnverifiedThe fundamental limitation of token-level scoring lies in the information bottleneck: aggregation happens after scoring, and fine-grained differences are already diluted before aggregation.64% similarUnverifiedThe paper points out that the analysis relies on moment conditions (p>1 or second moments), which heavy-tailed gradient distributions in deep learning may fail to satisfy.63% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1084113API
curl https://kongchang.com/api/v1/knowledge/claims/1084113MCP
get_claim(id=1084113)