MinerU-Diffusion
[ECCV 2026] A diffusion-based framework for document OCR that replaces autoregressive decoding with block-level parallel diffusion decoding.
- 624
- Python
- MIT
- 5 mo ago
[ECCV 2026] A diffusion-based framework for document OCR that replaces autoregressive decoding with block-level parallel diffusion decoding.
[ACL '26] UNCODE (a.k.a. PC-Sampler): training-free decoding calibration for masked diffusion LLMs. Code for "Empirical Analysis of Decoding Biases in Masked Diffusion Models"
Minimal masked diffusion LM in PyTorch. Pretraining + SFT + a scaling family from 50M to 350M, LLaDA recipe.
Text generation model, used with transformers.
Pure-CUDA inference engine + OpenAI-compatible server for LLaDA2.2-mini (16B MoE diffusion LLM) on Jetson Orin Nano Super 8GB
Text generation model, used with transformers.
Text generation model, used with transformers.
Text generation model, used with transformers.
Text generation model, used with transformers.