← Back to Videos

SGS: Contact-Rich Manipulation via Scaled Sim-to-Real RL

Loading video

Success-Guided Sampling trains 1x reactive policies with scaled-up sim-to-real RL: no demonstrations, retargeting, tactile sensors, or reward tuning, yet precise contact-rich manipulation on real hardware.

Sim-to-Real Reinforcement Learning Manipulation Contact-Rich

Category: manipulation

Author: @mateoguaman

Date: 2026-10-09T00:00:00

Duration: 7.168s

Source: https://x.com/mateoguaman/status/2108590647481397737

Reference: https://sgs-rl.github.io

Success-Guided Sampling: Mega-Scale RL for Contact-Rich Assembly

Success-Guided Sampling: Mega-Scale RL for Contact-Rich Assembly

@ctorobotics · 109 views · 2026-10-11

Success-Guided SamplingSGS-RLReinforcement Learning

Gemini 4 Argon Ties Opus 5.5 for SOTA on Real-Robot Manipulation

Gemini 4 Argon Ties Opus 5.5 for SOTA on Real-Robot Manipulation

@chooi_jeq · 77 views · 2026-10-14

RobocurveBenchmarksVLA

Self-Adaptive VLA: learning from its own failed rollouts to fix itself

Self-Adaptive VLA: learning from its own failed rollouts to fix itself

@icefox1104 · 739 views · 2026-09-28

VLASelf-Adaptive VLAPost-training

GEN-1.5 Runs a Continuous Cosmetics-Kit Assembly

GEN-1.5 Runs a Continuous Cosmetics-Kit Assembly

@GeneralistAI · 20 views · 2026-10-15

Robot foundation modelsDexterous ManipulationLong-Horizon Autonomy

UVTA: Unified Visual-Tactile-Action Modeling for Dexterous Manipulation

UVTA: Unified Visual-Tactile-Action Modeling for Dexterous Manipulation

@CyberRobooo · 217 views · 2026-10-10

Dexterous ManipulationTactile SensingHuman demonstrations

Stanford Real-Time EXPO-FT makes slow-but-smart VLAs real-time on robots: 42% to 97% with just 10 minutes of online data

Stanford Real-Time EXPO-FT makes slow-but-smart VLAs real-time on robots: 42% to 97% with just 10 minutes of online data

@LeoKharon · 482 views · 2026-10-02

VLAReinforcement LearningReal-time control

As an Amazon Associate, we earn from qualifying purchases.