Skip to content
RobotWorld
← Back to Videos

AlphaNIST: Binary Reward Plus Control Cost Hits 98.2% Sim Success

Loading video
Loading video

First post of the AlphaNIST thread: binary reward plus a control cost reaches 98.2% simulation success across 20 NIST assembly tasks, with no dense reward engineering.

Robotic ArmManipulationReinforcement LearningAssembly
Category: arm
Author: @octi_zhang
Date: 2026-10-13T00:00:00
Duration: 92.74s
Source: https://x.com/octi_zhang/status/2110084633161556468
Reference: https://arxiv.org/abs/2610.13354

Related Videos

AlphaNIST: One Policy From Scratch Solves All 20 NIST Assembly Tasks
0:07

AlphaNIST: One Policy From Scratch Solves All 20 NIST Assembly Tasks

@patrickhyin · 53 views · 2026-10-13
Robotic ArmManipulationReinforcement Learning
ACCRUE: Continual Off-Policy RL for Robot Foundation Models
1:00

ACCRUE: Continual Off-Policy RL for Robot Foundation Models

@cykim1006 · 33 views · 2026-10-14
Reinforcement LearningContinual learningRobot foundation models
AlphaNIST: Chaining Ten Robotic Assemblies from Sparse Rewards
0:10

AlphaNIST: Chaining Ten Robotic Assemblies from Sparse Rewards

@mateoguaman · 111 views · 2026-10-13
AlphaNISTReinforcement LearningRobotic Assembly
Pi API Enters Beta: Fine-Tune Pi Models for Customer Robots
0:40

Pi API Enters Beta: Fine-Tune Pi Models for Customer Robots

@chelseabfinn · 65 views · 2026-10-13
Robotic ArmDual-ArmPhysical Intelligence
Jev drives a MuJoCo metal arm with nine typed decisions
0:36

Jev drives a MuJoCo metal arm with nine typed decisions

@IsaacSin12 · 680 views · 2026-09-18
Robotic ArmManipulationMuJoCo
GPT-Policy: Dual-Arm Manipulation Learned In-Context from Demos
0:54

GPT-Policy: Dual-Arm Manipulation Learned In-Context from Demos

@z_code68632 · 865 views · 2026-09-15
In-Context LearningVLMManipulation

As an Amazon Associate, we earn from qualifying purchases.