DeepSeek-OCR · Issues· 288 open

Verify evals on Papers with Code

Improvement vectors for DeepSeek-OCR

基于中间层隐状态读取的内生安全通路——替代外挂护栏与参数对齐的第三种架构

prompt修改不起作用

[教程TUTORIAL] DeepSeek-OCR Windows WSL本地安装部署避坑指南 Local Installation and Deployment Guide for Windows WSL

Invitation: WDCC Shanghai subconference on the forgetting-mechanism direction

Does it support multiple images per prompt?

Hallucination/generation loop on Blackwell GPUs (sm_120)

【想法探讨】讨论是否可行--使用“文本内容（现代汉语）→文言文”转换极致压缩上下文输入

DeepSeek OCR Triton Error [CUDA] an illegal memory access on vLLM 0.11.2

表格识别的问题

能给出单个字符的矩形框吗（需要一个矩形框里最多一个字符，不能一个矩形框里多个字符）

Query About the Evaluation

Documentation Enhancement Suggestion

Is it only using 25% of the GPU for anyone else?