Congratulations to BAIR faculty @jlistgarten and @yun_s_song who have been elected to the International Society for Computational Biology (@iscb) 2026 class of Fellows!
iscb.org/about-iscb/soc…
English
Berkeley AI Research
1.4K posts
Berkeley AI Research
@berkeley_ai
We're graduate students, postdocs, faculty and scientists at the cutting edge of artificial intelligence research.
Berkeley, CA Joined Temmuz 2017
448 Following264.1K Followers
Congratulations to BAIR faculty @doristsao who has been elected to the American Academy of Arts and Sciences!
amacad.org/news/new-membe…
English
Berkeley AI Research retweeted
Heterogeneous simulation (different mesh per world) is now fully supported in mjlab. If you were hesitant about using mjlab for manipulation, now is a great time to switch over 🙂
mujocolab.github.io/mjlab/main/sou…
English
Berkeley AI Research retweeted
I trained a grasping policy on objects of various shapes and sizes using the new mjlab feature and this really cool pivot grasp strategy emerges for large flat objects that are larger than the gripper aperture. So cool and beautiful to see!
English
Berkeley AI Research retweeted
We wanted to make long-horizon video generation accessible beyond massive compute clusters. With Quant VideoGen, you can now deploy HYWorldPlay locally on an RTX 5090. By exploiting spatiotemporal redundancy, we’ve slashed memory usage by 7× with almost no overhead. Congratulations to the team!
Haocheng Xi@HaochengXiUCB
🎥 Video generation is hitting the memory wall. As videos get longer, the KV cache quietly explodes — and long-horizon consistency starts to break. We built Quant VideoGen: a training-free KV cache compression method for auto-regressive video diffusion. Instead of storing every KV in high precision, QVG exploits video’s spatiotemporal redundancy with semantic-aware smoothing + progressive residual quantization. 🚀 Up to 7× KV memory reduction ⚡ <4% overhead ✅ Strong long-video quality 🕹️ Deploy HYWorldPlay on your own RTX 5090 locally KV compression is becoming a core scaling primitive — not just for LLMs, but for video generation too. Paper: arxiv.org/abs/2602.02958 Code: github.com/svg-project/Qu… (1/5)
English
Berkeley AI Research retweeted
🎥 Video generation is hitting the memory wall.
As videos get longer, the KV cache quietly explodes — and long-horizon consistency starts to break.
We built Quant VideoGen: a training-free KV cache compression method for auto-regressive video diffusion.
Instead of storing every KV in high precision, QVG exploits video’s spatiotemporal redundancy with semantic-aware smoothing + progressive residual quantization.
🚀 Up to 7× KV memory reduction
⚡ <4% overhead
✅ Strong long-video quality
🕹️ Deploy HYWorldPlay on your own RTX 5090 locally
KV compression is becoming a core scaling primitive — not just for LLMs, but for video generation too.
Paper: arxiv.org/abs/2602.02958
Code: github.com/svg-project/Qu…
(1/5)
English
Berkeley AI Research retweeted
1/ "Can QPP Choose the Right Query Variant?" has been accepted at #SIGIR2026!🇦🇺
You can easily over-generate multiple query variants at low cost, but running RAG for all of them is expensive!
Can we pick the winner query before paying the generation cost?
arxiv.org/abs/2604.22661
English
Berkeley AI Research retweeted
Gave my PhD dissertation talk on Friday!
It's been an incredible journey made possible by the best advisor who believed in me and gave me the freedom and support to explore. Thank you @pabbeel!
And thank you to everyone who came to support and share this milestone with me 🙏
English
Berkeley AI Research retweeted
I am in Rio for #ICLR 2026! I am excited to be presenting 3 posters at the following workshops:
⚖️ Hierarchical Agenda Reasoning for Strategic Multi-Turn Dialogue Agents
📍 04/26, 11:15 AM–12 PM or 4:10–5 PM | DATA-FM, Room 203 A+B
📍 04/27, 11:35 AM–12:20 PM & 2:30–3:10 PM | SPOT Workshop
✏️ How LLMs Distort Our Written Language
📍 04/26, 3–3:50 PM | AIWILD Workshop, Room 204 A/B
🤖 Evaluating and Reducing Deceptive Dialogue from Language Models with Multi-Turn RL
📍 04/27, 10–11 AM | Trustworthy AI Workshop, Room 204 A+B
If you're interested in dialogue agents, multi-turn RL, AI deception, and/or preserving human agency in AI-interactions, I'd love to chat. Feel free to reach out!
English
Berkeley AI Research retweeted
Tomorrow in the Workshop on World Models at ICLR in Rio (10:30 am) I’ll talk about a… different take on what might make for a good world model. Come find out, 10:30 in Room 202A at ICLR sites.google.com/view/iclr-2026…
English
Berkeley AI Research retweeted
I will give two talks at ICLR workshops!! 🇧🇷
Sunday 9:40-10:10:
"LLMs for Distributed Data Use" @ Workshop on Data Problems in Foundation Models (Room 203 A/B)
Monday 15:30–16:05 :
"Are Mixture-of-Experts Modular? Why It Matters and How to Fix It" @ ICBINB Workshop (Room 201 C)
Both happened to be related to MoEs, but tackle two completely different questions → some say hi!
English
Berkeley AI Research retweeted
I'll give a talk about lifelong learning tmrw (Sun), 9:30 am (Brazil time) in the lifelong agents workshop, about how we can get robot foundation models to improve with RL: lifelongagent.github.io
Then at 11:30 am, I'll talk about how generative models can drive self-improvement here: recursive-workshop.github.io
English
Berkeley AI Research retweeted
What if we represent a state as a "list" of similarities to all other states? In our recent ICLR paper, we studied this "dual" representation.
Come visit our poster at #4608 10:30a-1p on Fri (morning, 2nd day)!
Paper: arxiv.org/abs/2510.06714
Blog post: seohong.me/blog/dual-repr…
Seohong Park@seohong_park
Introducing *dual representations*! tl;dr: We represent a state by the "set of similarities" to all other states. This dual perspective has lots of nice properties and practical benefits in RL. Blog post: seohong.me/blog/dual-repr… Paper: arxiv.org/abs/2510.06714 ↓
English
Berkeley AI Research retweeted
Thrilled to present GEPA as an Oral Talk and Poster at ICLR 2026 this Friday in Rio! 🇧🇷
Apr 24
Oral Session 3A (Agents), 10:30 AM BRT, Amphitheater
Poster Session 4, 3:15 PM, Pavilion 3
x.com/LakshyAAAgrawa…
Let's recap what's happened since we released GEPA last year 🧵
Lakshya A Agrawal@LakshyAAAgrawal
How does prompt optimization compare to RL algos like GRPO? GRPO needs 1000s of rollouts, but humans can learn from a few trials—by reflecting on what worked & what didn't. Meet GEPA: a reflective prompt optimizer that can outperform GRPO by up to 20% with 35x fewer rollouts!🧵
English
Berkeley AI Research retweeted
Excited to be in Rio this week to present RETAIN (w/ @zhiyuan_zhou_ , @ajwagenmaker, @KarlPertsch, and @svlevine) at #ICLR2026! 🇧🇷
Saturday 10:30 AM – 1:00 PM at P3-#1208.
Project Page: retain.yajatyadav.com
x.com/zhiyuan_zhou_/….
Paul Zhou@zhiyuan_zhou_
Do you ever find finetuning VLA overfits to the target task, to the point where generalist ability is lost and even minor deviations beyond the SFT data break the policy? We found an extremely simple solution: directly merge the base and finetuned policy in weight space 🤯 👇🧵
English
Berkeley AI Research retweeted
Excited to be in Rio attending ICLR this week to present some papers! 🧵(1/3)
(1) Decoupled Q-Chunking w/ @seohong_park and @svlevine on Fri 3:15-5:45 (#4504)
x.com/qiyang_li/stat…
Qiyang (Colin) Lil@qiyang_li
Action chunking is drawing growing interest in RL, yet its theoretical properties are still understudied. We are excited to share some insights on when we should use action chunking in Q-learning + a new algo (DQC) to tackle hard long-horizon tasks!colinqiyangli.github.io/dqc🧵1/N
English
Berkeley AI Research retweeted
🎉 Thrilled to have two papers accepted to ACL 2026 main!
1. Graph-based models match LLMs on close-ended human simulation tasks with far less compute & greater transparency
2. (oral) How to allocate human samples towards fine-tuning vs post-hoc rectification in simulation
English
Berkeley AI Research retweeted
Very excited to share this work @davidrmcall did with the fantastic NVIDIA Finland team last year. We have a surprisingly simple, but sample efficient way to post-train a flow model with RL.
David McAllister@davidrmcall
We developed a simple, sample-efficient online RL technique for post-training image generation models. We see it as a possible steerable alternative to CFG, driven by any scalar reward, including human preference.
English
Berkeley AI Research retweeted
🎉 The Agents in the Wild: Safety, Security, and Beyond workshop @ICLR2026 is less than a week away! Join us April 26 in Room 204 A/B, Riocentro, Rio de Janeiro! 🌴
Safety and security for AI agents — both foundational and emerging challenges — demand serious attention. Researchers and practitioners are mobilizing:
▪️ 151 papers accepted
▪️ 161 reviewers (58% industry, 42% academia)
▪️ Up to 800 participants expected
▪️ Incredible engagement on a topic that clearly matters.
The schedule: 👇
English
Berkeley AI Research retweeted
Optimize materials yourself! 🔥
No cloning, no install — just open our interactive Google Colab demo and start optimizing crystals for formation energy or band gap in < 2 minutes.
→ Try it here: colab.research.google.com/drive/1usUg7ze…
→ Or grab the repo + open weights: github.com/znowu/CliqueFl… huggingface.co/iamkuba/Clique…
English