SoMark 开源了 1.3B 参数量的文档解析模型 OXR，该模型在后训练中引入了递归自我 improvement 机制，登顶 olmOCR-Bench、OmniDocBench 和 DocParsingBench 三个基准榜单。OXR 能将 PDF 和文档图片转化为支持 Agent 的 Markdown 或 JSON 格式，其吞吐量达到了 GLM-OCR 的 3.05 倍。
OXR makes parsing failures part of the training process.
During post-training (OXR is the first document parser to use RSI here), a VLM evaluates its outputs, rollout disagreements flag uncertainty, and a separate clustering model groups the causes;
those findings then guide targeted data generation for the next round, and the 1.3B model that comes out of it sits first on olmOCR-Bench, OmniDocBench and DocParsingBench while running 3.05× GLM-OCR's throughput on a single RTX PRO 6000, the fastest open-source result in the speed comparison.
Worth studying (Github link in comment)
来源：Rohan Paul · x.com