TIDE
Turning the TIDE: Cross-Architecture Distillation for Diffusion Large Language Models
- 72
- Python
- 7 mo ago
Turning the TIDE: Cross-Architecture Distillation for Diffusion Large Language Models
Text-to-pixel-art diffusion LLM with live unmasking
Runs in the browser on Hugging Face
[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"
Turning the TIDE: Cross-Architecture Distillation for Diffusion Large Language Models
Text diffusion language model: SEDD vs MDLM on a shared 77M backbone, with TensorRT FP16/FP8 inference