Metraipolis › archive › dc_ecf29f18374a

DIAL-OPD: Learning More from Fewer Tokens in On-Policy Distillation — arXiv:2610.11659

Source
arXiv (en)
Original
https://arxiv.org/abs/2610.11659
License (as archived)
CC0-1.0
Archived
2026-10-09T06:00:34.533Z
SHA-256 of the archived text
30a25c5d3b0f2f031fa06c964c3e47a1ac1ad7e84d80b096dbe1985addc3dd47
Stable URL of this page
https://metraipolis.com/documento/dc_ecf29f18374a

How to cite

DIAL-OPD: Learning More from Fewer Tokens in On-Policy Distillation — arXiv:2610.11659 — arXiv (en), https://arxiv.org/abs/2610.11659 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_ecf29f18374a (Metraipolis archive dc_ecf29f18374a, sha256 30a25c5d3b0f2f031fa06c964c3e47a1ac1ad7e84d80b096dbe1985addc3dd47)
BibTeX
@misc{dcecf29f18374a,
  title = {DIAL-OPD: Learning More from Fewer Tokens in On-Policy Distillation — arXiv:2610.11659},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.11659},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_ecf29f18374a; sha256 30a25c5d3b0f2f031fa06c964c3e47a1ac1ad7e84d80b096dbe1985addc3dd47},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.11659. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

On-policy distillation (OPD) supervises student-generated trajectories with token-level teacher signals. Its sampled-token variant avoids the cost of full-vocabulary probabilities. Yet we find that training on fewer tokens can outperform full-token OPD, challenging the intuition that more supervision improves learning. This motivates selecting tokens by learning value. Existing disagreement-based criteria ignore probability scale: tokens assigned negligible probability by both models, termed low-low tokens, can receive large log-ratio rewards and hinder learning. We propose DIAL-OPD, a token-selection method that bridges log-probability and probability spaces by weighting reward magnitude with the logarithmic mean of teacher and student probabilities. A parameter beta controls this weighting, and the highest-scoring tokens are retained. Across 4 teacher-student pairs and 7 mathematical reasoning benchmarks, we compare DIAL-OPD with 9 baselines. Retaining only 40% of tokens, it outperforms Vanilla OPD and its full-token variants, with mean accuracy gains reaching 5.25 percentage points over Vanilla OPD, and doubles AIME25 Pass@16 from 13.33% to 26.67%. It also achieves up to an 18% relative improvement in mean accuracy over the strongest token-selection baseline at matched retention ratios. With a 4B teacher, DIAL-OPD surpasses the strongest full-token baseline using an 8B teacher at both student scales, showing that effective supervision allocation can outweigh teacher scaling. Further analysis shows that moderate beta balances suppressing low-low tokens against preserving useful disagreements. Token-level evidence reveals that DIAL-OPD filters high-reward tokens with limited reasoning value while preserving supervision critical to reasoning correctness.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno).
Categorie arXiv: Computation and Language (cs.CL) ; Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2610.11659 — 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 2026-10-09T05:47:01.812Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).