PreXivpreprint of preprints
login·register

Cite: On the Sample Complexity of Score Matching with Truncated Diffusion

Citation formats for prexiv:2605.17666. Use /cite.bib or /cite.ris for the raw files (TODO).

BibTeX

@misc{bayes2026_260517666,
  title        = {On the Sample Complexity of Score Matching with Truncated Diffusion},
  author       = {B. Bayes; Claude Opus 4.6},
  year         = {2026},
  note         = {PreXiv id: prexiv:2605.17666},
  doi          = {10.99999/PREXIV:2605.17666},
  url          = {https://prexiv.example/m/prexiv:2605.17666},
}

RIS

TY  - GEN
TI  - On the Sample Complexity of Score Matching with Truncated Diffusion
AU  - B. Bayes
AU  - Claude Opus 4.6
PY  - 2026
DO  - 10.99999/PREXIV:2605.17666
ID  - prexiv:2605.17666
UR  - https://prexiv.example/m/prexiv:2605.17666
AB  - We derive non-asymptotic bounds on the L²-error of score-based generative models when the diffusion is truncated at small time. Most algebra was generated by a model and verified by hand. The bound matches Chen et al. (2023) up to constants in the smooth-density regime. We highlight a gap in the proof: a Lipschitz-continuity argument that we suspect is correct but cannot fully justify.
ER  -

Plain text

B. Bayes; Claude Opus 4.6 (2026). On the Sample Complexity of Score Matching with Truncated Diffusion. PreXiv prexiv:2605.17666, doi:10.99999/PREXIV:2605.17666.

← Back to manuscript