Seth Z. Zhao

CS PhD Student @ UCLA | Prev @ UC Berkeley | Qualcomm/Vals AI Fellow

[Seth Z. Zhao @sethzhao506]

It would be interesting to see how it performs under current LLM’s training recipes of pre training, mid training and post training stage. I think this is like a discovery problem that you would like the agent to find the best solution for your optimizer in training time under different model/task settings, but for such discovery problem normally requires different domains methods to see the effectiveness.

[jiagengliu @jiagengliu02]

🚀 Introducing 𝐀𝐮𝐭𝐨𝐀𝐠𝐞𝐧𝐭𝟎: An 𝐀𝐠𝐞𝐧𝐭𝐢𝐜 𝐑𝐮𝐧𝐭𝐢𝐦𝐞 for Safe Closed-Loop Driving!

Most autonomous driving research focuses on optimizing E2E driving policies. However, in long-tail, 𝐥𝐚𝐫𝐠𝐞𝐥𝐲 𝐮𝐧𝐯𝐞𝐫𝐢𝐟𝐢𝐚𝐛𝐥𝐞 driving environments, stronger policies alone cannot ensure 𝐬𝐚𝐟𝐞 𝐚𝐧𝐝 𝐫𝐨𝐛𝐮𝐬𝐭 𝐜𝐥𝐨𝐬𝐞𝐝-𝐥𝐨𝐨𝐩 𝐞𝐱𝐞𝐜𝐮𝐭𝐢𝐨𝐧.

In 𝐀𝐮𝐭𝐨𝐀𝐠𝐞𝐧𝐭𝟎, we reformulate traditional runtime monitoring as an 𝐚𝐠𝐞𝐧𝐭𝐢𝐜 𝐫𝐮𝐧𝐭𝐢𝐦𝐞 𝐡𝐚𝐫𝐧𝐞𝐬𝐬 that reasons about 𝐰𝐡𝐞𝐧 𝐭𝐨 𝐢𝐧𝐭𝐞𝐫𝐯𝐞𝐧𝐞 and 𝐡𝐨𝐰 𝐭𝐨 𝐫𝐞𝐜𝐨𝐯𝐞𝐫, leveraging two key capabilities:

1️⃣ 𝐒𝐜𝐞𝐧𝐞 𝐂𝐨𝐧𝐭𝐞𝐱𝐭 𝐑𝐞𝐟𝐢𝐧𝐞𝐦𝐞𝐧𝐭 — Leveraging VLM reasoning to refine critical scene context missed by nominal policies, enabling reliable action verification and triggering interventions when nominal policies fail.

2️⃣ 𝐀𝐠𝐞𝐧𝐭𝐢𝐜 𝐑𝐞𝐜𝐨𝐯𝐞𝐫𝐲 𝐋𝐨𝐨𝐩 — Leveraging VLM-guided stochastic optimization to iteratively propose, verify, and refine explainable recovery policies through verifier feedback.

📊 𝐒𝐭𝐫𝐨𝐧𝐠 𝐜𝐥𝐨𝐬𝐞𝐝-𝐥𝐨𝐨𝐩 𝐠𝐚𝐢𝐧𝐬! 𝐀𝐮𝐭𝐨𝐀𝐠𝐞𝐧𝐭𝟎 consistently improves representative E2E driving policies on 𝐅𝐚𝐢𝐥𝟐𝐃𝐫𝐢𝐯𝐞 (up to +37.0 SR) and 𝐍𝐚𝐯𝐒𝐚𝐟𝐞-∞ (up to +32.8 SR), without policy retraining.

Our results highlight 𝐚𝐠𝐞𝐧𝐭𝐢𝐜 𝐫𝐮𝐧𝐭𝐢𝐦𝐞 𝐚𝐬 𝐚 𝐜𝐨𝐦𝐩𝐥𝐞𝐦𝐞𝐧𝐭𝐚𝐫𝐲 𝐝𝐢𝐫𝐞𝐜𝐭𝐢𝐨𝐧 𝐭𝐨 𝐬𝐜𝐚𝐥𝐢𝐧𝐠 𝐄𝟐𝐄 𝐝𝐫𝐢𝐯𝐢𝐧𝐠 𝐩𝐨𝐥𝐢𝐜𝐢𝐞𝐬.

🎉 Huge congrats to the amazing team: @jiagengliu02 , @sethzhao506 , @Zewei__Zhou , Alejandro Huerta, Aidan Bayer-Calvert, Yuxin Bao, @DarthUtopian , Robert Spataru, @zijun_cui02 , @yining_hong , @zhoubolei , and Jiaqi Ma!

🛠️ Everything is open-sourced!

🌐 Project: https://t.co/N7KMJZndiQ
⭐ Code: https://t.co/VXTVXlNeRA
📄 Paper: https://t.co/AdbYhXU2Em

#AgenticAI #AutonomousDriving #EndToEndDriving #PhysicalAI

[Yuxin Bao @YuxinBc13752hz]

Introducing NavSafe-∞: an open, photorealistic CL benchmark with 280 event-based scenarios across 28 safety-critical event types.

Across 20 E2E policies, we expose the OL-to-CL safety gap.

Project, paper & code: https://t.co/WYpjh60fSh

(1/5)

[Vals AI @ValsAI]

We are thrilled to introduce the first cohort of Vals Fellows!

These five researchers are building evals that tackle some of AI’s most important questions: from model safety to formal mathematical reasoning, as well as real world engineering and security challenges.

Get to know the researchers building the next generation of evals in the thread below.

[Seth Z. Zhao @sethzhao506]

@ValsAI @a16z @8vc @pearvc @BloombergBeta @HRTVentures @nextladder Congrats!

[Seth Z. Zhao @sethzhao506]

@RayanKrishnan This might be a signal of subliminal learning https://t.co/yq2lhep70o transmitted during distillation

[Vals AI @ValsAI]

Meta just released Muse Spark 1.1 and is the new SOTA on MedScribe and TaxEval, taking the top spot from Fable 5 while being 10x cheaper and twice as fast. Meta currently holds the top 2 spots on TaxEval

It is also the new #1 on Harvey's Legal Agent Bench, dethroning Grok 4.5 less than 24 hours after it took the top spot.

[Qualcomm Research & Technologies @QCOMResearch]

Congrats to the Qualcomm Innovation Fellowship NA winners 👏
PhD research + mentorship from @Qualcomm engineers = real-world impact.
See what they’re building:
https://t.co/I5e1ONUyNk

[Chenfeng_X @Chenfeng_X]

Excited that our paper StreamdiffusionV2 received the Best Research Paper Award at #MLSys26!

🚀Video generation is quickly moving from demos to production-facing workloads. It is no longer a turn-based pipeline but should be a streaming pipeline to interact with users.

📖Our project page: https://t.co/ItuO5zc6hT and paper: https://t.co/fmz2irYIm1

👂Come join the talk if you are interested in streaming video generation. Our talk will be at the Research Track Oral Presentation: Best Paper Session on Tue 8:45AM at #MLSys26 , I will talk about how we attacked the efficiency and quality challenges. Hope to see you there!

❤️Huge thanks to all authors! This work would not have been possible without the incredible effort from the entire team. Big shout out to Tianrui Feng, Zhi Li, @Andy_ShuoYang , @HaochengXiUCB, @lmxyy1999 , @lvminzhang , @xiuyu_l , Keting Yang, @ZiqiPeng, @songhan_mit , @magrawala, @KurtKeutzer , and @cumulo_autumn

[Hongyang Li @francislee2020]

Introducing #WorldEngine, https://t.co/UK3vsulS2u, a two-year long project. The missing infrastructure for Physical AI post-training in Autonomous driving. Open-source. Production-validated.

[Zewei Zhou @Zewei__Zhou]

Thrilled to share that our paper 𝐀𝐮𝐭𝐨𝐕𝐋𝐀 has been accepted to 𝐍𝐞𝐮𝐫𝐈𝐏𝐒 𝟐𝟎𝟐𝟓! See you in San Diego! 🦭

🗓️ Dec. 3 (Wed), 11:00 AM – 2:00 PM PST
📍 Exhibit Hall C,D,E # 2208

[Andrew Ng @AndrewYNg]

Releasing a new "Agentic Reviewer" for research papers. I started coding this as a weekend project, and @jyx_su made it much better.

I was inspired by a student who had a paper rejected 6 times over 3 years. Their feedback loop -- waiting ~6 months for feedback each time -- was painfully slow. We wanted to see if an agentic workflow can help researchers iterate faster.

When we trained the system on ICLR 2025 reviews and measured Spearman correlation (higher is better) on the test set:
- Correlation between two human reviewers: 0.41
- Correlation between AI and a human reviewer: 0.42

This suggests agentic reviewing is approaching human-level performance.

The agent grounds its feedback by searching arXiv, so it works best in fields like AI where research is freely published there. It’s an experimental tool, but I hope it helps you with your research.

Check it out here: https://t.co/n7ctnDilJJ

[Sunday @sundayrobotics]

After 18 months in stealth, dozens of prototypes, millions of real-home demonstrations, and one final all-nighter, we’re thrilled for you to say hello to Memo

[Sunday @sundayrobotics]

November 19

[Zewei Zhou @Zewei__Zhou]

Excited to share my 𝐭𝐰𝐨 𝐟𝐢𝐫𝐬𝐭-𝐚𝐮𝐭𝐡𝐨𝐫 𝐩𝐚𝐩𝐞𝐫𝐬 accepted at 𝐈𝐂𝐂𝐕'𝟐𝟓!🌺

🤔 Can multi-agent collaboration move us beyond single-frame perception toward spatio-temporal scenario understanding, facilitating end-to-end optimization across multiple driving tasks?

[Chris Huang @ZhiyuChris]

🚗Join us at @ICCVConference 2025 Tutorial!
📌Beyond Self-Driving: Exploring Three Levels of Driving Automation
📅Oct 19 (Sun), 8:50 – 12:10
📍Room 308A
Discover how self-driving is evolving from single-vehicle intelligence to connected, cooperative, and city-scale systems.

[Chenfeng_X @Chenfeng_X]

📢 Excited to sharing a little late update (before it is no longer news): I’ll be joining @UTAustin @UTCompSci as an Assistant Professor! I'm recruiting PhD students from @UTCompSci in the Fall 2025 cycle and also looking for RAs/interns! More info see https://t.co/JPDhVplhJX

[Seth Z. Zhao @sethzhao506]

Huge thanks to the workshop organizers and my advisors and co-authors: Hao Xiang, @Chenfeng_X , Xin Xia, @zhoubolei, Jiaqi Ma.

[Seth Z. Zhao @sethzhao506]

It's my pleasure to share that our project CooPre: Cooperative Pretraining for V2X Cooperative Perception has been accepted by #IROS2025 and awarded Best Paper at the #CVPR2025 DriveX Workshop (https://t.co/Mk4DfAx0D9)!
Paper could be found at: https://t.co/5lM4xXjJPM

[Seth Z. Zhao @sethzhao506]

CooPre tackles a major challenge in V2X Cooperative Perception by introducing a multi-agent self-supervised pretraining framework that enhances data efficiency in multi-agent systems.