AI 精选动态
智能评分 60
百度推出Unlimited-OCR新模型
AI 推荐理由
Unlimited-OCR通过R-SWA实现高效长文档解析，对比DeepSeek-OCR有明确量化优势，值得关注OCR技术进展。核心解读
百度PP-OCR团队发布Unlimited-OCR模型，采用R-SWA注意力机制保持KV缓存恒定，支持单次处理数百页PDF，在OmniDocBench基准上达到93%准确率，相较DeepSeek-OCR提升6%。
全文
Baidu Inc. (@Baidu_Inc) 转发了 jzhang533 (@jzhang533) 的帖子：
Baidu’s open OCR stack:
PP-OCR: fast text detection + recognition
PaddleOCR-VL: turns pages into structured docs
Qianfan-OCR: promptable document intelligence
Unlimited-OCR: long-horizon PDF/multi-page parsing
OCR is documents becoming computable.
> **引用原帖 Adina Yakup (@AdinaYakup):**
> Unlimited-OCR 🔥New OCR from @PaddlePaddle
> It can parse hundreds of pages in a single pass while maintaining stable speed.
> The key idea is R-SWA (Reference Sliding Window Attention), which keeps KV cache constant during decoding.
> 🏆 93% on OmniDocBench
> 📈 +6% over DeepSeek-OCR
> https://x.com/AdinaYakup/status/2069081754837614998