Twitxy Saved

Jaemoo Choi

@JaemooChoi

Georgia Tech
Atlanta, GA jaemoo-choi.github.io Joined June 2025
126 Following 209 Followers 42 Posts

Latest posts 5

Pinned post

Jaemoo Choi @JaemooChoi
Simple, efficient, scalable diffusion RL without likelihood ratios?

Meet RVM: Reward-based Velocity Matching

🎯 Rewards directly in velocity space
⚡ Cheap video fine-tuning
🧩 Unifies recent objectives
🎬 Fixes the static-video problem by dynamic motion reward

How it works 🧵
0:21
1 replies 18 reposts 107 likes 14K views View on X
Jaemoo Choi @JaemooChoi
Excited to share our work, MDraft! 🚀

We present a new observation that leads to a new training objective, substantially strengthening the drafter and substantially improves draft acceptance rates!
Petr @pmolodyk Nov 1
[1/8] Can we improve parallel drafters with on-policy distillation?
Introducing MDraft: on-policy distillation for diffusion drafters.
At T=1, mean acceptance length vs. DFlash:
• Qwen3-4B: +11.9%
• Qwen3-8B: +13.4%
Same inference, just swap the drafter checkpoint! 🧵
View quoted post on X
0 replies 0 reposts 16 likes 649 views View on X
Jaemoo Choi @JaemooChoi
I’ll be at ICML in Seoul 🇰🇷!

I’m mainly interested in RL post-training, self-improving AI, long-context / long-horizon generation, and improving autoregressive video models.

If you’d like to exchange ideas or grab a coffee, feel free to shoot me a DM!
4 replies 5 reposts 65 likes 4.6K views View on X
Jaemoo Choi @JaemooChoi
Thrilled to announce that "Rethinking Design space of RL for Diffusion Models" has been accepted at ICML 2026, see you in 🇰🇷!
Jaemoo Choi @JaemooChoi Mar 6
We proudly present “Rethinking the Design Space of RL for Diffusion Models”

showing that ELBO-based likelihood estimation (from the final sample) is the dominant driver of stable, efficient RL fine-tuning. On SD3.5-Medium, we boost GenEval 0.24 → 0.95 in ~90 GPU hours, beating FlowGRPO (4.6×) and DiffusionNFT (2×) efficiency.

Great Collab with @YongxinChen1 @YuchenZhu_ZYC @WeiGuo01 @MoleiTaoMath Petr Molodyk, Bo Yuan, Jinbin Bai, Yi Xin

🎥 Video attached
📍Link: arxiv.org/abs/2602.04663

#DiffusionModels #ReinforcementLearning #TextToImage #GenAI
View quoted post on X
2 replies 7 reposts 62 likes 6.5K views View on X
Jaemoo Choi @JaemooChoi
We developed a full extension of target matching (adjoint matching) and Adjoint Schrödinger Bridge Sampler (ASBS) for discrete domain! 🚀

Key contributions in the paper:
🌉 Extended Schrödinger Bridge (SB) theory to the discrete domain.
🔄Developed adjoint matching and ASBS for discrete domains by leveraging cyclic group structures.
Wei Guo @WeiGuo01 Mar 16
How to bring the speed & precision of continuous adjoint matching (AM) to discrete neural samplers?
Introducing discrete adjoint Schrödinger bridge sampler (DASBS): a unified framework for authentic discrete AM! 🎲✨
Joint work with @JaemooChoi et al.:
📄arxiv.org/abs/2602.08243
View quoted post on X