Metraipolis › archive › dc_0ca7d2e5cba5

Noise Your Prompt: Noising Conditioning Tokens in Continuous Diffusion Language Models — arXiv:2610.09145

Source
arXiv (en)
Original
https://arxiv.org/abs/2610.09145
License (as archived)
CC0-1.0
Archived
2026-10-08T11:26:29.840Z
SHA-256 of the archived text
b65201ddb30dc80b7883df5f489fb994e6b947a1cc9d6215a8f38174afa9093b
Stable URL of this page
https://metraipolis.com/documento/dc_0ca7d2e5cba5

How to cite

Noise Your Prompt: Noising Conditioning Tokens in Continuous Diffusion Language Models — arXiv:2610.09145 — arXiv (en), https://arxiv.org/abs/2610.09145 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_0ca7d2e5cba5 (Metraipolis archive dc_0ca7d2e5cba5, sha256 b65201ddb30dc80b7883df5f489fb994e6b947a1cc9d6215a8f38174afa9093b)
BibTeX
@misc{dc0ca7d2e5cba5,
  title = {Noise Your Prompt: Noising Conditioning Tokens in Continuous Diffusion Language Models — arXiv:2610.09145},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.09145},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_0ca7d2e5cba5; sha256 b65201ddb30dc80b7883df5f489fb994e6b947a1cc9d6215a8f38174afa9093b},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.09145. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

We revisit a standard accepted practice in the continuous diffusion language model literature of fixing conditioning prompt tokens clean during training. We make a very simple modification: also noise the conditioning prompt tokens during training. We demonstrate that under this modified training objective, we achieve better generalization in combinatorial reasoning tasks such as Sudoku and N-Queens, with the largest gains on harder variants ($3.73\% \to 24.65\%$ solve rate on Sudoku Hard), and increased diversity of generated solutions ($50.60\% \to 73.79\%$ coverage on 10x10 N-Queens). We also show measurable improvements to natural language generation quality in modest dataset regimes with Gigaword summarization, but notably demonstrate that gains do not transfer to all natural language tasks (e.g open ended dialogue generation). Our method is a single line change to the training objective, requires no additional inference costs by default, and provides the flexibility of classifier-free guidance inspired guided sampling. Our \href{ this https URL } {code} is publicly available.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.CL, nuovi lavori del giorno).
Categorie arXiv: Machine Learning (cs.LG) ; Computation and Language (cs.CL).
Pagina del lavoro: https://arxiv.org/abs/2610.09145 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-08T10:24:06.366Z dalla pagina https://arxiv.org/list/cs.CL/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).