←
SISL509 posts
SISL avatar

SISL

@SISLaboratory

Stanford Intelligent Systems Laboratory, Aeronautics & Astronautics Department. Advancing Research on Autonomous Systems and Decision Making Under Uncertainty.

◷ Joined September 2015▧ 177 Photos and videos
Max Lamparth avatar
⟳ SISL retweeted
Max Lamparth@MLamparth·1h

Really enjoyed discussing with @farhatchristina what it takes to build AI that acts with us and for us, from trust and human oversight to independent evaluations and policy. Check it out!

Hoover Research Fellow @MLamparth joins Stanford MIP student @FarhatChristina for a discussion on when AI can be trusted to inform consequential decisions, and where responsibility should remain when AI agents act or advise. Watch Nerds of the Roundtable: hoover.org/events/ai-trust-a…

SISL avatar
SISL@SISLaboratory·Nov 12

Exciting news! SISL director @aiprof_mykel is joining Stanford Center for International Security and Cooperation (CISAC) 🙌

Please join us in welcoming @aiprof_mykel to CISAC! Kochenderfer, an associate professor of aeronautics and astronautics at Stanford School of Engineering, will serve as CISAC's natural science co-director beginning January 1, 2027. cisac.fsi.stanford.edu/news/…

SISL avatar
SISL@SISLaboratory·Nov 9

Try it out: pip install retriever-core (Apache 2.0) or check it out here: retriever.systems/ arxiv.org/abs/2607.17213 Work by @LinfengZhaoZLF, @HaojieHuang13, @maojiayuan, @Weiyu_Liu_, Lawson Wong, and @aiprof_mykel, a collaboration between Stanford, Northeastern, and MIT.

SISL avatar
SISL@SISLaboratory·Nov 9

Retriever-0, the demonstration agent built on top of it, searches drawers for black pepper, seasons a steak, and retrieves and sorts items from a deformable bag using bimanual skills, all from a natural language goal.

SISL avatar
SISL@SISLaboratory·Nov 9

Retriever is a new programming model from SISL for building such systems. Each component, whether a camera, a memory of past observations, a vision-language planner, or a learned skill, is written as a reusable Python module with its own clock. Modules compose like PyTorch.

Post attachment
SISL avatar
SISL@SISLaboratory·Nov 9

In practice, that means wiring together perception, memory, planners, learned skills, and controllers that each run at their own pace, and the glue code is usually the least principled part of the system.

SISL avatar
SISL@SISLaboratory·Nov 9

New SISL paper and tool! Most robot demos work because the robot can see everything it needs, but real environments are not like that. Robots have to act to gather information, keep track of what they have already checked, and updates their plans, even for simple retrieval tasks.

Post attachment
SISL avatar
SISL@SISLaboratory·Oct 22

Now online! Recording of Marc Schlichting’s PhD defense (Nov 15, 2025) on generative models for safety in autonomous systems, from failure discovery in a 1200-dim F-16 system to an LLM decision-support tool for aviation emergencies tested in real flights. youtube.com/watch?v=nCtkqVG7…

Post attachment
SISL avatar
SISL@SISLaboratory·Oct 1

New work by SISL'er @houjun_liu. Check it out!

🚨new method day! 🚨 Trust[ing results] in ML conferences is utterly broken. Let's fix it with an algorithm! We are excited to introduce 𝗧𝗿𝗮𝗶𝗻𝗶𝗻𝗴 𝗪𝗶𝘁𝗻𝗲𝘀𝘀𝗲𝘀, a system for “honesty” certificates of data and evaluations which a verifier can cheaply check and trust.

Post attachment
SISL avatar
SISL@SISLaboratory·Sep 14

We evaluate on a masked-pixel benchmark and on a real-world intrusion discovery task using indirect muon tomography, with improved estimation accuracy and planning efficiency over conventional methods.

Post attachment