Detect, Remask, Repair: Diffusion Editing for Faithful Summarization of Evolving Contexts
Quick Answer
This paper shows that The DETECT-REMASK-REPAIR framework utilizes diffusion models to update outdated summaries while preserving supported content, achieving significant efficiency with repair times under half a second.
Quick Take
Evaluated on DialogSum and the new StreamSum benchmark, it demonstrates improved faithfulness and controllability in evolving-context summarization.
Key Points
- Localized faithfulness repair updates summaries without full regeneration.
- Repair times reduced to under half a second using one-step repair.
- StreamSum benchmark introduced for evaluating evolving-context summarization.
- Faithfulness-steered repair improves early drafts significantly.
- Framework allows trade-offs between faithfulness, speed, and preservation.
Paper Resources
Source Excerpt
arXiv:2606. 12807v1 Announce Type: new Abstract: Summaries of real-world events can become outdated as contexts evolve and new information arrives.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 We study localized faithfulness repair: updating outdated spans in an existing summary while preserving supported content.
We propose DETECT-REMASK-REPAIR, a diffusion-based framework that identifies, remasks, and repairs outdated regions with masked diffusion language models. …
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from arXiv cs.CL
See more →TriAgent: Divergence-Aware Committees for Cost-Efficient Financial Sentiment Analysis
TriAgent introduces a cost-efficient multi-agent system for financial sentiment analysis, combining VADER, FinBERT, and Qwen2.5. It achieves an F1 score of ~0.87 with significant savings of $9.3M/year at a 10M-user scale compared to GPT-4o-mini, while also detecting hallucinations with an AUC of 0.90.