# @kepano I just tried it this morning on the 245-page Mythos pdf and it failed badly and the outputs ...

Canonical URL: https://www.traeai.com/articles/49163e11-b4d2-4cc3-bc10-a71731330c7a
Original source: https://x.com/karpathy/status/2042292197287215230
Source name: Andrej Karpathy(@karpathy)
Content type: tweet
Language: 英文
Score: 5.5
Reading time: 2 分钟
Published: 2026-04-09T17:22:58+00:00
Tags: 大语言模型, PDF处理, AI工程

## Summary

Andrej Karpathy指出当前SOTA大模型在处理245页PDF时表现糟糕，认为PDF转换应作为LLM的专用技能而非通用程序。

## Key Takeaways

- 当前顶尖大模型处理复杂PDF文档仍存在严重问题
- PDF内容提取失败表现为输出混乱、结构错乱
- 建议将PDF解析设计为LLM的专项技能而非通用功能

## Citation Guidance

When citing this item, prefer the canonical traeai article URL for the AI-readable summary and include the original source URL when discussing the underlying source material.