Metraipolis › archive › dc_32d33e507aac

Look Before You Select: Rethinking Vocabulary Sparsification in On-Policy Distillation — arXiv:2609.34386

Source: arXiv (en)
Original: https://arxiv.org/abs/2609.34386
License (as archived): CC0-1.0
Archived: 2026-10-07T21:53:24.224Z
SHA-256 of the archived text: fe588e8117008744629a37bbaadd3336a9e88eaada559a5408903730244024ad
Stable URL of this page: https://metraipolis.com/documento/dc_32d33e507aac

How to cite

Look Before You Select: Rethinking Vocabulary Sparsification in On-Policy Distillation — arXiv:2609.34386 — arXiv (en), https://arxiv.org/abs/2609.34386 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_32d33e507aac (Metraipolis archive dc_32d33e507aac, sha256 fe588e8117008744629a37bbaadd3336a9e88eaada559a5408903730244024ad)

BibTeX

@misc{dc32d33e507aac,
  title = {Look Before You Select: Rethinking Vocabulary Sparsification in On-Policy Distillation — arXiv:2609.34386},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.34386},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_32d33e507aac; sha256 fe588e8117008744629a37bbaadd3336a9e88eaada559a5408903730244024ad},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.34386. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

On-policy distillation (OPD) uses teacher correction on student-generated responses. Full-vocabulary correction can provide important corrections even for tokens that the student assigns low probability, but backpropagating through all token logits becomes memory-intensive for long sequences. Existing memory-saving approaches estimate corrections from sampled tokens or restrict supervision to the student's TopK tokens, introducing sampling noise or changing the full-vocabulary correction. We introduce \textbf{SparseOPD}, which uses full-vocabulary teacher correction to determine which corrections matter before selecting the token logits to differentiate. SparseOPD first constructs the full-vocabulary correction without retaining its backward graph, then selects tokens by correction magnitude rather than student probability. Signed residual compensation preserves the total promoting and suppressing correction mass, while correction-aware budget allocation distributes the sparse support across positions. Finally, the update backpropagates only through the selected token logits. Across six task--scale settings spanning mathematics, chemistry QA, and multimodal reasoning, SparseOPD outperforms Sampled Token and TopK in task-average accuracy and matches or exceeds Full Vocabulary. Gradient cosine similarity reaches 99\% on 4B mathematics, while 8K full-parameter profiling shows 70.5\% lower backward memory.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-29 in cs.CL).
Categorie arXiv: Computation and Language (cs.CL) ; Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2609.34386 — 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 2026-10-07T21:27:08.610Z dalla pagina https://arxiv.org/catchup/cs.CL/2026-09-29?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).