yuezhouhu/residual-context-diffusion
Residual Context Diffusion (RCD): Repurposing discarded signals as structured priors for high-performance reasoning in dLLMs.
This project offers an advanced way to improve how quickly and accurately large language models (LLMs) can solve complex problems, especially in mathematics. It takes mathematical problems as input and produces more precise answers faster than existing methods by intelligently reusing discarded computational information. Data scientists, machine learning engineers, and researchers working with advanced LLMs for reasoning tasks would find this particularly useful.
Use this if you need to significantly boost the accuracy and parallel processing speed of diffusion-based LLMs for demanding mathematical or reasoning challenges.
Not ideal if your primary need is for basic chat-style LLM interactions or if you are not working with diffusion LLMs specifically.
Stars
55
Forks
2
Language
Python
License
—
Category
Last pushed
Mar 12, 2026
Commits (30d)
0
Get this data via API
curl "https://pt-edge.onrender.com/api/v1/quality/diffusion/yuezhouhu/residual-context-diffusion"
Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.
Higher-rated alternatives
ljleb/sd-mecha
Executable State Dict Recipes
SJTU-DENG-Lab/Discrete-Diffusion-Forcing
Discrete Diffusion Forcing (D2F): dLLMs Can Do Faster-Than-AR Inference
declare-lab/tango
A family of diffusion models for text-to-audio generation.
Li-Jinsong/DAEDAL
[ICLR 2026] Official repository of "Beyond Fixed: Training-Free Variable-Length Denoising for...
SalesforceAIResearch/CoDA
Salesforce AI Research's open diffusion language model