{"slug": "detect-remask-repair-diffusion-editing-for-faithful-summarization-of-evolving", "title": "Detect, Remask, Repair: Diffusion Editing for Faithful Summarization of Evolving Contexts", "summary": "A new diffusion-based framework called DETECT-REMASK-REPAIR updates outdated spans in existing summaries while preserving supported content, offering a localized alternative to full regeneration. The approach, evaluated on the StreamSum benchmark of synthetic event timelines, improves faithfulness in early drafts and reduces repair cost to under half a second for one-step repairs. The framework also serves as a post-hoc correction step that enhances faithfulness for autoregressive summarization systems.", "body_md": "arXiv:2606.12807v1 Announce Type: new\nAbstract: Summaries of real-world events can become outdated as contexts evolve and new information arrives. A common response is to generate a new summary from the updated context, but full regeneration discards the previous draft, can obscure what changed, and may be unnecessary when only a few claims are unsupported. We study localized faithfulness repair: updating outdated spans in an existing summary while preserving supported content. We propose DETECT-REMASK-REPAIR, a diffusion-based framework that identifies, remasks, and repairs outdated regions with masked diffusion language models. To evaluate evolving-context summarization, we introduce StreamSum, a benchmark of synthetic event timelines. Experiments on DialogSum and StreamSum show that localized diffusion repair provides a controllable alternative to full rewriting: faithfulness-steered repair improves early drafts, one-step repair reduces repair cost to under half a second, with the framework enabling faithfulness-speed-preservation tradeoffs across datasets. We also find that the framework can provide a post-hoc correction step that improves faithfulness for autoregressive systems.", "url": "https://wpnews.pro/news/detect-remask-repair-diffusion-editing-for-faithful-summarization-of-evolving", "canonical_source": "https://arxiv.org/abs/2606.12807", "published_at": "2026-06-12 04:00:00+00:00", "updated_at": "2026-06-12 04:56:02.984728+00:00", "lang": "en", "topics": ["artificial-intelligence", "natural-language-processing", "generative-ai", "large-language-models", "ai-research"], "entities": ["DETECT-REMASK-REPAIR", "StreamSum", "DialogSum"], "alternates": {"html": "https://wpnews.pro/news/detect-remask-repair-diffusion-editing-for-faithful-summarization-of-evolving", "markdown": "https://wpnews.pro/news/detect-remask-repair-diffusion-editing-for-faithful-summarization-of-evolving.md", "text": "https://wpnews.pro/news/detect-remask-repair-diffusion-editing-for-faithful-summarization-of-evolving.txt", "jsonld": "https://wpnews.pro/news/detect-remask-repair-diffusion-editing-for-faithful-summarization-of-evolving.jsonld"}}