Steering Away from Memorization: Reachability-Constrained Reinforcement Learning for Text-to-Image Diffusion
Paper • 2603.00140 • Published
How to use S-Karnik/rads-model-checkpoints with Diffusers:
pip install -U diffusers transformers accelerate
import torch
from diffusers import DiffusionPipeline
# switch to "mps" for apple devices
pipe = DiffusionPipeline.from_pretrained("S-Karnik/rads-model-checkpoints", dtype=torch.bfloat16, device_map="cuda")
prompt = "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k"
image = pipe(prompt).images[0]Steering Away from Memorization: Reachability-Constrained Reinforcement Learning for Text-to-Image Diffusion (NeurIPS 2026)
Sathwik Karnik*, Juyeop Kim*, Sanmi Koyejo, Jong-Seok Lee, Somil Bansal
Stanford University · Yonsei University (*equal contribution)
Paper · Code · Project Page
@inproceedings{karnik2026rads,
title = {Steering Away from Memorization: Reachability-Constrained Reinforcement Learning for Text-to-Image Diffusion},
author = {Karnik, Sathwik and Kim, Juyeop and Koyejo, Sanmi and Lee, Jong-Seok and Bansal, Somil},
booktitle = {Advances in Neural Information Processing Systems (NeurIPS)},
year = {2026}
}