Collected molecules will appear here. Add from search or explore.
A 1B-parameter diffusion-based small language model trained on FineWeb, utilizing the LLaDA framework for bidirectional context and denoising-based text generation.
stars
0
forks
0
This is a very early-stage project (7 days old, 0 stars) attempting to replicate or apply the LLaDA (Large Language Diffusion with Autoregression) architecture. While diffusion for text is an interesting niche compared to standard autoregressive models, this specific project lacks the community traction, unique dataset, or architectural breakthrough required for a higher defensibility score. Frontier labs are actively researching discrete diffusion for LLMs, posing a high risk of obsolescence.
TECH STACK
INTEGRATION
library_import
READINESS