Baidu's Unlimited OCR Gen 2 Ingests 3,600-Page Harry Potter Series in One Pass
xiaohu · x · 2026-10-27
Baidu's Wenxin Unlimited OCR second generation pushes document parsing from tens of pages toward "unlimited". It introduces the Unlimited-Horizon Decoding Paradigm, enabling end-to-end ingestion of thousands of pages in one pass — demonstrated on the 1,200+ page Three-Body Problem trilogy and the 3,600+ page Harry Potter series.
Key mechanism: like a human copyist who only focuses on the current lines rather than memorizing the whole book, the model:
- splits documents into page blocks sharing a boundary page;
- after finishing a block, keeps the boundary page's visual KV and recent text KV, reuses them in the next block, and frees stale page caches;
- so only the active block's cache resides on GPU, keeping attention cost bounded.
Related event: Baidu Open-Sources Unlimited OCR 2 for Thousand-Page Document Parsing(5 posts)→
More from Models
- Dev flags 'concerning' Cohere open-model behavior: system-wide scan and data upload to US entity — basedjensen · 2026-10-27
- Developer banned after using Codex on his own logs; mark_k asks OpenAI to investigate — mark_k · 2026-10-27
- Even Heavy Users Keep Forgetting Haiku and Sonnet 5.5 Exist — repligate · 2026-10-27
- Tagging @bot on X automatically forwards your post to Grok Bot — EricBuess · 2026-10-27
- Smaug Humanizer goes open-weights: de-AI-slop your text, built on Gemma — bindureddy · 2026-10-27
- Model Lineups Defended: 'One Benchmark Score Can't Capture a Model's Utility' — craigsdennis · 2026-10-27