{"slug": "rewrite-then-veto-rewriter-agnostic-edit-gating-for-llm-based-post-ocr", "title": "Rewrite, Then Veto: Rewriter-Agnostic Edit Gating for LLM-Based Post-OCR Correction", "summary": "A 0.6B-parameter veto gate called EditVeto improves LLM-based post-OCR correction across all five rewriters tested, including two unseen in training, according to an arXiv paper (2611.05539v1). On the eight HIPE-OCRepair-2026 test sets, the gate cut weighted cMER by 0.0026-0.0044 and raised the preference score by 0.15-0.25 (paired bootstrap, p ~ 0.0002 each), though on published competition outputs it lowered cMER for three of four teams but raised it for the winning run, from 0.0050 to 0.0065. A retrained gate in a second disclosed test look still improved LoRA-tuned rewriters (cMER -0.0011 to -0.0021), and an ensemble of their gated edits reached a cMER of 0.0069 and a preference score of 0.928, mostly through fine-tuning and pooling rather than gating.", "body_md": "arXiv:2611.05539v1 Announce Type: new \nAbstract: LLMs can repair noisy OCR transcripts of historical documents, but they also change text that was already correct. We propose EditVeto, a veto gate placed after a full-context LLM rewrite. The rewrite is character-aligned to the OCR input, its differences are split into blocks, and a 0.6B-parameter gate keeps or reverts each block. An oracle analysis motivates the design: given the true error locations, per-span generation is worse than no correction, whereas vetoing a full rewrite's edits helps. Trained on rewrites by two Qwen3 models and evaluated, frozen, on the eight HIPE-OCRepair-2026 test sets, the gate improves all five rewriters tested, two of them unseen in training: weighted cMER falls by 0.0026-0.0044 and the preference score rises by 0.15-0.25 (paired bootstrap, p ~ 0.0002 each). On published competition outputs it lowers cMER for three of four teams but raises it for the winning run (0.0050 to 0.0065). In a second, disclosed test look, a retrained gate still improves LoRA-tuned rewriters (cMER -0.0011 to -0.0021), and an ensemble of their gated edits reaches a cMER of 0.0069 and a preference score of 0.928, mostly through fine-tuning and pooling rather than gating. Built only on public models and data, it is behind the winner (0.0050) and tied with the second-ranked run (0.0071), both from the only team using domain pre-training, and ahead of all other published runs. A third, secondary look shows the unchanged gate improving zero-shot Qwen3.8-27B (cMER 0.0114 to 0.0108) but only the preference score of its LoRA-tuned version.", "url": "https://wpnews.pro/news/rewrite-then-veto-rewriter-agnostic-edit-gating-for-llm-based-post-ocr", "canonical_source": "https://arxiv.org/abs/2611.05539", "published_at": "2026-11-05 05:00:00+00:00", "updated_at": "2026-11-05 05:19:27.594103+00:00", "lang": "en", "topics": ["large-language-models", "ai-research", "natural-language-processing", "machine-learning"], "entities": ["EditVeto", "HIPE-OCRepair-2026", "Qwen3", "Qwen3.8-27B", "arXiv"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/rewrite-then-veto-rewriter-agnostic-edit-gating-for-llm-based-post-ocr", "markdown": "https://wpnews.pro/news/rewrite-then-veto-rewriter-agnostic-edit-gating-for-llm-based-post-ocr.md", "text": "https://wpnews.pro/news/rewrite-then-veto-rewriter-agnostic-edit-gating-for-llm-based-post-ocr.txt", "jsonld": "https://wpnews.pro/news/rewrite-then-veto-rewriter-agnostic-edit-gating-for-llm-based-post-ocr.jsonld"}}