Infinity-Parser2 model advances document parsing with synthetic data and multi-task RL

Researchers have introduced Infinity-Parser2, a large multimodal model designed for end-to-end document parsing. The model utilizes a controllable data-synthesis pipeline and multi-task reinforcement learning to overcome the scarcity of annotated parsing data. Key contributions include the creation of Infinity-Doc2-5M, a 5-million-sample bilingual corpus, and a novel reward system for joint reinforcement learning across eight co-trained objectives. Two variants, Infinity-Parser2-Flash and Infinity-Parser2-Pro, have been released, with the latter achieving state-of-the-art results on benchmarks like olmOCR-Bench and ParseBench. AI

IMPACT Advances document parsing capabilities, potentially improving efficiency and accuracy in handling diverse document types.

RANK_REASON The cluster describes a technical report detailing a new multimodal model and dataset released on arXiv. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• arXiv
• DeepSeek-OCR-2
• Hugging Face
• Infinity-Doc2-5M
• Infinity-Parser2
• Infinity-Parser2-Flash
• Infinity-Parser2-Pro
• Infinity-Parser-7B
• MinerU2.5
• olmOCR-Bench
• PaddleOCR-VL-1.5
• ParseBench

• paper
• model release
• infra

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →