Metraipolis › archive › dc_9da25e0bc4de

A Heckler in the Hidden State: Correctness Signals in Diffusion Language Models — arXiv:2609.36783

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.36783
License (as archived)
CC0-1.0
Archived
2026-10-07T17:59:03.776Z
SHA-256 of the archived text
7a1995cf774cca720314ea272e7ed1ef962cc4f1caf68a61222cb373ef8bae76
Stable URL of this page
https://metraipolis.com/documento/dc_9da25e0bc4de

How to cite

A Heckler in the Hidden State: Correctness Signals in Diffusion Language Models — arXiv:2609.36783 — arXiv (en), https://arxiv.org/abs/2609.36783 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_9da25e0bc4de (Metraipolis archive dc_9da25e0bc4de, sha256 7a1995cf774cca720314ea272e7ed1ef962cc4f1caf68a61222cb373ef8bae76)
BibTeX
@misc{dc9da25e0bc4de,
  title = {A Heckler in the Hidden State: Correctness Signals in Diffusion Language Models — arXiv:2609.36783},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.36783},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_9da25e0bc4de; sha256 7a1995cf774cca720314ea272e7ed1ef962cc4f1caf68a61222cb373ef8bae76},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.36783. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Diffusion language models generate code by repeatedly updating a partially masked sequence. We ask whether their internal activations encode code correctness and whether that information can improve generation. Across six diffusion models, linear probes distinguish passing from failing attempts, with the strongest reads generally appearing beyond the early layers. Controls using small semantic mutations support a connection to correctness rather than surface style alone. In comparisons with model confidence, probe point estimates offer no consistent advantage. Adding a probe-derived direction to the residual stream does not yield a dependable improvement in the tested steering settings, while the opposite direction degrades performance. We distinguish these observations from claims about statistical significance or a general inability to steer. Supplementary methods, archived results, and code document the tested interventions and the limits of their statistical calibration and reproducibility.

Tema degli esploratori: calcolo e infrastrutture (retro-catalogo arXiv: annunci del 2026-09-30 in cs.SE).
Categorie arXiv: Software Engineering (cs.SE).
Pagina del lavoro: https://arxiv.org/abs/2609.36783 — 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 2026-10-07T17:31:52.004Z dalla pagina https://arxiv.org/catchup/cs.SE/2026-09-30?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).