Metraipolis › archive › dc_9846f905343e

Summarize, Judge, Refine: Decoupled Content Understanding and Policy Learning for Multimodal Content Moderation — arXiv:2609.22094

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.22094
License (as archived)
CC0-1.0
Archived
2026-10-09T06:48:46.164Z
SHA-256 of the archived text
86fda4836dbaac196f97b818ffc27a823a445844f3d030e2fd3a1570e5d1b383
Stable URL of this page
https://metraipolis.com/documento/dc_9846f905343e

How to cite

Summarize, Judge, Refine: Decoupled Content Understanding and Policy Learning for Multimodal Content Moderation — arXiv:2609.22094 — arXiv (en), https://arxiv.org/abs/2609.22094 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_9846f905343e (Metraipolis archive dc_9846f905343e, sha256 86fda4836dbaac196f97b818ffc27a823a445844f3d030e2fd3a1570e5d1b383)
BibTeX
@misc{dc9846f905343e,
  title = {Summarize, Judge, Refine: Decoupled Content Understanding and Policy Learning for Multimodal Content Moderation — arXiv:2609.22094},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.22094},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_9846f905343e; sha256 86fda4836dbaac196f97b818ffc27a823a445844f3d030e2fd3a1570e5d1b383},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.22094. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Content moderation systems traditionally entangle multimodal understanding with policy-specific classification, requiring full pipeline retraining for every policy change and suffering from label scarcity since multimedia cannot be meaningfully augmented. We propose Summarize-Judge-Refine (SJR), a two-model architecture that decouples these concerns via a natural language interface: a multimodal Content Model produces structured text summaries, and a text-only Policy Model classifies them against policy definitions. An iterative co-training loop refines the Content Model via GRPO to produce policy-relevant summaries, while text-space augmentation generates adversarial summary variants---an augmentation pathway impossible on raw multimedia---enabling few-shot policy bootstrap. Every decision is grounded in a human-readable summary, providing interpretability as a structural byproduct. On misleading advertisement detection, SJR achieves +23.6\% relative non-misleading F1 over a zero-shot chain-of-thought baseline, outperforming end-to-end SFT, STaR/RFT, and RLFT. Notably, a variant trained on zero real violating examples---with all positive-class data synthetically generated---matches the full-data model within 0.2\% relative on violating F1, demonstrating that new policies can launch without any real violation data.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-22 in cs.CL).
Categorie arXiv: Computation and Language (cs.CL) ; Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2609.22094 — 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 2026-10-09T06:48:44.815Z dalla pagina https://arxiv.org/catchup/cs.CL/2026-09-22?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).