Unverified50% confidenceFactExact time
The paper covers two settings, compactly supported targets and power-law targets, under the p-Wasserstein distance framework, gives non-asymptotic convergence bounds, and separates the contributions of early-stopping error, initialization error, score approximation error, and time discretization error.
1
Sources
50%
Confidence
Long-term
Relevance
10/15/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedCoarse-grained error type classification relies more on semantic understanding, while fine-grained error span localization relies more on boundary judgment and position awareness, and the model's strengths and weaknesses on these two types of tasks are very likely not consistent.72% similarUnverifiedThe study explicitly decomposes the approximation error into two parts—spectral truncation error and circuit approximation error—making the overall approximation rate a quantifiable and controllable result.70% similarUnverifiedThe paper establishes non-asymptotic, high-probability upper bounds for the convergence of empirical occupation measures to the invariant measure, under the contraction assumption in kernel-induced geometry and the standard spectral gap assumption.70% similarUnverifiedThe study establishes non-asymptotic bounds on the conditional score estimation error, i.e., bounds that hold for finite sample sizes70% similarUnverifiedThe error bound in this paper holds for all Wasserstein-p distances with p ≥ 1, rather than being limited to a specific metric.70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1037351API
curl https://kongchang.com/api/v1/knowledge/claims/1037351MCP
get_claim(id=1037351)