AI Paper Atlas

Research map / Reinforcement learning / RL from feedback

Iterative Distillation for Reward-Guided Fine-Tuning of Diffusion Models in Biomolecular Design

Xingyu Su, Xiner Li, Masatoshi Uehara, Sunwoo Kim, Yulai Zhao, Gabriele Scalia, Ehsan Hajiramezanali, Tommaso Biancalani, Degui Zhi, Shuiwang Ji

ICLR 2026 · Reinforcement learning › RL from feedback · approach: RL fine-tuning of generative models

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Use VIDD instead of DDPO (PPO-style)diffusion fine-tuning · IFNAR2 binder designipTM: 0.2403 -> 0.5090Improved
Use VIDD instead of DDPOEvoDiff · Protein SS-matchpLDDT: 0.55 -> 0.82Improved
Use VIDD instead of DRAKES (reward backprop)DNA diffusion · DNA Enhancer HepG2Pred-Activity: 6.44 -> 8.28Improved
Add Best-of-N sampling to VIDD (ablation)VIDD · Molecule Docking Parp1docking score: 9.4 -> 12.1Improved
Add Best-of-N (N=128) to VIDD (ablation)VIDD · IFNAR2 binder designipTM: 0.51 -> 0.66Improved
Update roll-out policy every step (K=1) rather than K=5 (ablation)VIDD · DNA Enhancer HepG2Pred-Activity: 8.28 -> 7.06Hurt
Lazy update interval K=50 rather than 5 (ablation)VIDD · DNA Enhancer HepG2Pred-Activity: 8.28 -> 7.23Hurt
Increase regularization alpha from 1.0 to 2.0 (ablation)VIDD · DNA Enhancer HepG2Pred-Activity: 8.28 -> 7.26Hurt
Decrease alpha from 1.0 to 0.8 (ablation)VIDD · DNA Enhancer HepG2Pred-Activity: 8.28 -> 8.21No clear effect
Use VIDD instead of DDPOdiffusion fine-tuning · DNA Enhancer HepG23-mer Corr: 0.398 -> 0.162Hurt
Use VIDD instead of DDPOdiffusion fine-tuning · Protein SS-matchDiversity: 0.52 -> 0.52No clear effect

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy