{"version":"1.0","generated_at":"2026-10-13T01:19:08.102587","id":1653,"slug":"baidu-s-unlimited-ocr-processes-dozens-of-document-pages-in-one-pass-by-treating-memory-like-human-forgetting","title":"百度“无限OCR”通过将记忆模拟为人类遗忘，单次处理数十页文档","summary":"百度研究人员开发了Unlimited OCR模型，通过参考滑动窗口注意力（R-SWA）机制，实现单次推理处理数十页文档，内存和速度保持恒定，在基准测试中准确率达93%。","abstract":"百度“无限OCR”通过将记忆模拟为人类遗忘，单次处理数十页文档 百度研究人员开发了Unlimited OCR模型，通过参考滑动窗口注意力（R-SWA）机制，实现单次推理处理数十页文档，内存和速度保持恒定，在基准测试中准确率达93%。 百度研究人员开发了Unlimited OCR模型，通过参考滑动窗口注意力（R-SWA）机制，实现单次推理处理数十页文档，内存和速度保持恒定，在基准测试中准确率达93%。 原贴提到：Baidu researchers have built an OCR model that handles dozens of documen 来源：the-decoder.com 百度研究人员构建了一个OCR模型，单次推理即可处理数十页文档，无论文本长度如何，内存使用和速度都保持恒定。重新设计的注意力机制使之成为可能。百度研究人员在其技术报告中写道，目前没有OCR模型能在单次推理中处理超过大约十页。瓶颈是KV缓存，这是一个缓冲区，语言模型在生成过程中存储所有先前处理的标记，以便以后查找。当前的端到端系统使用语言模型作为解码器，因此这个缓冲区会随着每一行新文本而增长。这导致内存使用增加，生成速度逐渐变慢。实践中，系统通过一个循环逐页处理每个文档，每一步后重置缓存来绕开这个问题。 百度用一个人类类比来描述这个问题。抄写书籍的人不会重读他们已经写下的所有内容。他们盯着原文、刚刚写下的最后几个字符以及下一个要写的字符。较早的段落通过一种软性遗忘逐渐淡出。研究人员希望Unlimited OCR模仿这种模式。它通过团队所称的参考滑动窗口注意力（R-SWA）来实现。每个生成的标记…","access_level":"member","access_label":"免费阅读","access_mode":"full","is_preview":false,"canonical_url":"https://opc.beizhux.com/content/1653/baidu-s-unlimited-ocr-processes-dozens-of-document-pages-in-one-pass-by-treating-memory-like-human-forgetting","html_url":"https://opc.beizhux.com/content/1653/baidu-s-unlimited-ocr-processes-dozens-of-document-pages-in-one-pass-by-treating-memory-like-human-forgetting","json_url":"https://opc.beizhux.com/content/1653/baidu-s-unlimited-ocr-processes-dozens-of-document-pages-in-one-pass-by-treating-memory-like-human-forgetting.json","published_at":"2026-07-05T23:25:07","updated_at":"2026-10-13T00:23:15","category":{"slug":"ai-business","name":"AI小生意项目库"},"source":{"site":"the-decoder.com","author":"Jonathan Kemper","url":"https://the-decoder.com/baidus-unlimited-ocr-processes-dozens-of-document-pages-in-one-pass-by-treating-memory-like-human-forgetting/"},"tags":["AI","AI技术","AI推理","AI效率提升","AI模型","Deepseek","KV缓存","OCR","R-SWA","The Decoder","效率提升","文档处理"],"topics":[{"slug":"ai-side-business","name":"AI副业","url":"https://opc.beizhux.com/topics/ai-side-business","reason":"这篇内容命中「项目、小生意、变现」等主题信号。"},{"slug":"ai-super-individual","name":"AI超级个体","url":"https://opc.beizhux.com/topics/ai-super-individual","reason":"这篇内容命中「效率、学习」等主题信号。"},{"slug":"ai-tools","name":"AI工具","url":"https://opc.beizhux.com/topics/ai-tools","reason":"这篇内容命中「模型」等主题信号。"}],"keywords":["AI小生意项目库","AI副业","AI超级个体","AI工具","工具","自动化","模型","Cursor","Claude","工作流","副业","创业"],"questions":[{"question":"百度“无限OCR”通过将记忆模拟为人类遗忘，单次处理数十页文档主要讲什么？","answer":"百度研究人员开发了Unlimited OCR模型，通过参考滑动窗口注意力（R-SWA）机制，实现单次推理处理数十页文档，内存和速度保持恒定，在基准测试中准确率达93%。"},{"question":"这篇文章最值得关注的要点是什么？","answer":"百度研究人员开发了Unlimited OCR模型，通过参考滑动窗口注意力（R-SWA）机制，实现单次推理处理数十页文档，内存和速度保持恒定，在基准测试中准确率达93%。；原贴提到：Baidu researchers have built an OCR model that handles dozens of documen；来源：the-decoder.com"},{"question":"这篇文章和哪些AI专题相关？","answer":"它适合放在AI副业、AI超级个体、AI工具专题里阅读。 关联原因：这篇内容命中「项目、小生意、变现」等主题信号。；这篇内容命中「效率、学习」等主题信号。；这篇内容命中「模型」等主题信号。"},{"question":"阅读这篇文章建议先理解哪些关键词？","answer":"建议先理解AI工具、工具、自动化、模型、Cursor这些关键词，再结合正文判断工具、机会或风险是否值得进入自己的工作流。"}],"terms":[{"slug":"ai-tools-term","name":"AI工具","definition":"在AI觉醒星球里，「AI工具」属于「AI工具」方向。围绕AI工具、模型能力、自动化流程和真实使用场景做整理，优先关注能提升效率、降低成本和创造新交付的工具。 不只看工具热度，更看它能不能进入真实流程。","topic_slug":"ai-tools","topic_name":"AI工具","topic_title":"AI工具：模型、插件、自动化与实战场景","topic_url":"https://opc.beizhux.com/topics/ai-tools","topic_path":"/topics/ai-tools","url":"https://opc.beizhux.com/glossary/ai-tools-term","path":"/glossary/ai-tools-term","json_url":"https://opc.beizhux.com/glossary/ai-tools-term.json"},{"slug":"tools","name":"工具","definition":"在AI觉醒星球里，「工具」属于「AI工具」方向。围绕AI工具、模型能力、自动化流程和真实使用场景做整理，优先关注能提升效率、降低成本和创造新交付的工具。 不只看工具热度，更看它能不能进入真实流程。","topic_slug":"ai-tools","topic_name":"AI工具","topic_title":"AI工具：模型、插件、自动化与实战场景","topic_url":"https://opc.beizhux.com/topics/ai-tools","topic_path":"/topics/ai-tools","url":"https://opc.beizhux.com/glossary/tools","path":"/glossary/tools","json_url":"https://opc.beizhux.com/glossary/tools.json"},{"slug":"automation","name":"自动化","definition":"在AI觉醒星球里，「自动化」属于「AI工具」方向。围绕AI工具、模型能力、自动化流程和真实使用场景做整理，优先关注能提升效率、降低成本和创造新交付的工具。 不只看工具热度，更看它能不能进入真实流程。","topic_slug":"ai-tools","topic_name":"AI工具","topic_title":"AI工具：模型、插件、自动化与实战场景","topic_url":"https://opc.beizhux.com/topics/ai-tools","topic_path":"/topics/ai-tools","url":"https://opc.beizhux.com/glossary/automation","path":"/glossary/automation","json_url":"https://opc.beizhux.com/glossary/automation.json"},{"slug":"ai-model","name":"模型","definition":"在AI觉醒星球里，「模型」属于「AI工具」方向。围绕AI工具、模型能力、自动化流程和真实使用场景做整理，优先关注能提升效率、降低成本和创造新交付的工具。 不只看工具热度，更看它能不能进入真实流程。","topic_slug":"ai-tools","topic_name":"AI工具","topic_title":"AI工具：模型、插件、自动化与实战场景","topic_url":"https://opc.beizhux.com/topics/ai-tools","topic_path":"/topics/ai-tools","url":"https://opc.beizhux.com/glossary/ai-model","path":"/glossary/ai-model","json_url":"https://opc.beizhux.com/glossary/ai-model.json"},{"slug":"cursor","name":"Cursor","definition":"在AI觉醒星球里，「Cursor」属于「AI工具」方向。围绕AI工具、模型能力、自动化流程和真实使用场景做整理，优先关注能提升效率、降低成本和创造新交付的工具。 不只看工具热度，更看它能不能进入真实流程。","topic_slug":"ai-tools","topic_name":"AI工具","topic_title":"AI工具：模型、插件、自动化与实战场景","topic_url":"https://opc.beizhux.com/topics/ai-tools","topic_path":"/topics/ai-tools","url":"https://opc.beizhux.com/glossary/cursor","path":"/glossary/cursor","json_url":"https://opc.beizhux.com/glossary/cursor.json"},{"slug":"claude","name":"Claude","definition":"在AI觉醒星球里，「Claude」属于「AI工具」方向。围绕AI工具、模型能力、自动化流程和真实使用场景做整理，优先关注能提升效率、降低成本和创造新交付的工具。 不只看工具热度，更看它能不能进入真实流程。","topic_slug":"ai-tools","topic_name":"AI工具","topic_title":"AI工具：模型、插件、自动化与实战场景","topic_url":"https://opc.beizhux.com/topics/ai-tools","topic_path":"/topics/ai-tools","url":"https://opc.beizhux.com/glossary/claude","path":"/glossary/claude","json_url":"https://opc.beizhux.com/glossary/claude.json"},{"slug":"workflow","name":"工作流","definition":"在AI觉醒星球里，「工作流」属于「Agent工作流」方向。聚合Agent、AI编程、Claude Code、Cursor、工作流自动化等内容，帮助你把智能体能力落到真实任务链路。 把智能体当成任务系统，而不是聊天窗口。","topic_slug":"agent-workflow","topic_name":"Agent工作流","topic_title":"Agent工作流：智能体、AI编程与自动化协作","topic_url":"https://opc.beizhux.com/topics/agent-workflow","topic_path":"/topics/agent-workflow","url":"https://opc.beizhux.com/glossary/workflow","path":"/glossary/workflow","json_url":"https://opc.beizhux.com/glossary/workflow.json"},{"slug":"side-business","name":"副业","definition":"在AI觉醒星球里，「副业」属于「AI副业」方向。筛选AI副业、AI小生意、项目认知和可收费交付案例，关注普通人可以验证、复制和沉淀的方法。 从真实痛点出发，先做能收费的小验证。","topic_slug":"ai-side-business","topic_name":"AI副业","topic_title":"AI副业：小生意、项目机会与可收费交付","topic_url":"https://opc.beizhux.com/topics/ai-side-business","topic_path":"/topics/ai-side-business","url":"https://opc.beizhux.com/glossary/side-business","path":"/glossary/side-business","json_url":"https://opc.beizhux.com/glossary/side-business.json"}],"preview_hidden_sections":[],"related_items":[{"id":3365,"title":"Grok Bot 离线误报问题已修复","url":"https://opc.beizhux.com/content/3365/grok-bot","summary":"Grok Bot 的离线误报问题已修复，修复包含在 0.70.2 版本中，并正在推送。","access_level":"member"},{"id":3366,"title":"华为拟将 HarmonyOS 推向全球市场","url":"https://opc.beizhux.com/content/3366/harmonyos","summary":"余承东称华为正考虑逐步将 HarmonyOS 推向全球市场，但计划仍处初步阶段；同时披露超 9000 万台设备运行 HarmonyOS 6/7、2026 年底目标 1 亿台、生态含超 45 万应用服务与 1100 万注册开发者。","access_level":"member"},{"id":3360,"title":"AI 智能体团队消耗大量 token，质量提升却微乎其微，研究显示","url":"https://opc.beizhux.com/content/3360/ai-agent-teams-waste-massive-tokens-for-barely-measurable-quality-gains-research-finds","summary":"Vals AI 等测试显示，多智能体团队比单智能体贵 1.8–5.1 倍，四次比较仅一次显著提升；最大推理下几乎无优势。Noam Brown 称多智能体主要买速度，Eric Provencher 称协调税导致浪费。","access_level":"member"}],"usage_policy":{"summarizable":true,"preferred_url":"https://opc.beizhux.com/content/1653/baidu-s-unlimited-ocr-processes-dozens-of-document-pages-in-one-pass-by-treating-memory-like-human-forgetting","private_fields_excluded":true}}