From c479c1ad7222faed4fdb914625a139324a29b72f Mon Sep 17 00:00:00 2001 From: Ronald Tse Date: Fri, 4 Sep 2026 12:25:25 +0200 Subject: [PATCH] =?UTF-8?q?docs:=20Paper=20A=20references=20=E2=80=94=20ve?= =?UTF-8?q?rified=20pool=20+=20honest=20camera-ready=20flags?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Citations reused from Paper B's verified list (Sadeed/SadeedDiac-25, WikiNews-2024 multi-ref, ByT5/T5, distillation lineage, Interscript registry); vendor tech reports for LLM rows and the Semitic aux-task literature are explicitly flagged for the camera-ready pass rather than reconstructed from memory. --- docs/paper-a.adoc | 36 +++++++++++++++++++++++++++++------- 1 file changed, 29 insertions(+), 7 deletions(-) diff --git a/docs/paper-a.adoc b/docs/paper-a.adoc index d5f505e..55af4fe 100644 --- a/docs/paper-a.adoc +++ b/docs/paper-a.adoc @@ -101,13 +101,6 @@ we are not aware of a published single-variable comparison of the auxiliary's representation under fixed everything-else, which is the gap Section 4 fills. -[NOTE] --- -Full bibliographic entries are assembled in the camera-ready pass from -the rababa literature notes; the numbers above are the measured/ -published values our leaderboard rows carry. --- - == 3. Protocol === 3.1 Windowed DER-CE @@ -268,6 +261,35 @@ ledger (RESULTS.md) with every verdict, interval, and provenance hash. Paper B carries the client-tier artifacts (distilled IMF v1 zips, parity and margin reports, cross-runtime byte parity). +== References + +* [%hardbreaks] + Aldallal, Z., Chrouf, S., Hennara, K., Hamed, M. M., Hreden, M., AlModhayan, S. _Sadeed: Advancing Arabic diacritization through small language model._ arXiv:2504.21635, 2025. (Introduces the Sadeed-1.5B baseline; the SadeedDiac-25 benchmark protocol.) + + +* Mohamed, A., Mubarak, H. _Advancing Arabic diacritization: Improved datasets, benchmarking, and state-of-the-art models._ EMNLP 2025, pp. 16718-16730. (WikiNews-2024 multi-reference evaluation.) + + +* Xue, L. et al. _ByT5: Towards a token-free future with pre-trained byte-to-byte models._ NeurIPS 2021. + + +* Raffel, C. et al. _Exploring the limits of transfer learning with a unified text-to-text transformer._ JMLR 2020. + + +* Hinton, G., Vinyals, O., Dean, J. _Distilling the knowledge in a neural network._ arXiv:1503.02531, 2015. + + +* Kim, Y., Rush, A. _Sequence-level knowledge distillation._ EMNLP 2016. + + +* Tse, R. et al. _Interscript: an interoperable script conversion system registry._ 2022. + +[NOTE] +-- +LLM leaderboard rows (Claude-3.7-Sonnet, GLM-5.2/5.3, Gemini-Flash-2.0, +GPT-4) are identified by their vendors' published model identifiers and +measured values under the protocol of Section 3; per-vendor technical +reports are cited in the camera-ready pass. The auxiliary-supervision +literature comparison (Section 2) draws on the Semitic multi-task +morphology line catalogued in the companion repository's literature +notes; its per-work citations are likewise completed at camera-ready +from those notes rather than from memory. +-- + == Allocation note Paper A per the 2026-08-29 two-paper decision (PUBLICATION-NOTES.md):