Metraipolis › archive › dc_8dc517dea8a6

A Unified Per-Token Gating Family for On-Policy Distillation: FKL/RKL Mixing with Multi-Channel and Bias Coefficients — arXiv:2609.11768

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.11768
License (as archived)
CC0-1.0
Archived
2026-10-16T05:24:30.698Z
SHA-256 of the archived text
6021b3ffc229fe3afc13171cd5b6a77439782d31bc8e94d61968f39be2d9d81e
Stable URL of this page
https://metraipolis.com/documento/dc_8dc517dea8a6

How to cite

A Unified Per-Token Gating Family for On-Policy Distillation: FKL/RKL Mixing with Multi-Channel and Bias Coefficients — arXiv:2609.11768 — arXiv (en), https://arxiv.org/abs/2609.11768 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_8dc517dea8a6 (Metraipolis archive dc_8dc517dea8a6, sha256 6021b3ffc229fe3afc13171cd5b6a77439782d31bc8e94d61968f39be2d9d81e)
BibTeX
@misc{dc8dc517dea8a6,
  title = {A Unified Per-Token Gating Family for On-Policy Distillation: FKL/RKL Mixing with Multi-Channel and Bias Coefficients — arXiv:2609.11768},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.11768},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_8dc517dea8a6; sha256 6021b3ffc229fe3afc13171cd5b6a77439782d31bc8e94d61968f39be2d9d81e},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.11768. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Per-token gating of forward/reverse KL losses has become a standard technique for on-policy knowledge distillation (OPD), but existing methods such as EOPD (Jin et al., 2026) and ToDi (Jung et al., 2025) each fix a single gating signal and a single gating direction, and the two have never been compared directly. We introduce a four-coefficient parameterization lambda_t = sigma(a * h_t + b * u(x) + c + d * gap_t) in which direction-aligned proxies of EOPD and ToDi appear as one-dimensional (1D) restrictions, and which adds multi-channel composition and an explicit bias as further degrees of freedom. On TweetEval (Barbieri et al., 2020) emotion and hate, with a Qwen3-32B teacher and a Qwen3-4B student, configurations in the full family reach higher accuracy than the matched-magnitude single-channel (entropy-only / gap-only) 1D restrictions in 33 of 36 comparable cells, and a 26-cell mean-match isolation experiment places dynamic gating ahead of effective-KL-matched static baselines in 19 of 26 cells. Because cells share training data, models, and parameter substructure, we report both counts as exploratory aggregate directional evidence rather than as independent hypothesis tests. Targeted three-seed paired replications of the nine headline comparisons singled out by that sweep -- including a third task, offensive -- are directionally consistent, but individually smaller than the single-seed estimates and not significant at n=3. We therefore present the parameterization primarily as a shared coordinate system for comparing per-token gating designs in short-output classification OPD.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-11 in cs.AI).
Categorie arXiv: Artificial Intelligence (cs.AI) ; Computation and Language (cs.CL); Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2609.11768 — 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 2026-10-16T05:17:52.051Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-11?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).