Skip to content
RobotWorld
← Back

The Imitator Game: Benchmarking Robot Imitation Beyond Action Prediction

Loading video
Loading video

The Imitator Game benchmark progressively alters the scene relative to human demos to test whether robots grasp intent or just copy motions. All 9 SOTA models stay below 13% zero-shot success on unseen tasks; IG-10K provides 20k+ paired human-robot episodes across 50+ tasks.

Imitator Game模仿学习imitation learning具身智能embodied AIVLA基准测试benchmarkIG-10K人类视频模仿robot learningTwitter
Category: research
Author: @heetezition
Date: 2026-08-31T00:00:00
Duration: 171.502s
Source: https://x.com/heetezition/status/2094330599167451271
Reference: https://arxiv.org/abs/2608.22301

Related Videos

Workshop: making real-world RL practical
0:45

Workshop: making real-world RL practical

@yigitkkorkmaz · 62 views · 2026-08-31
强化学习RLreinforcement learning
VLAct: Representation-Centric Continued Pre-training for VLA Models
0:28

VLAct: Representation-Centric Continued Pre-training for VLA Models

@HuggingPapers · 21 views · 2026-08-31
VLActVLA视觉-语言-动作模型
Galaxea G0.5: One Autoregressive VLA for Reasoning and Action
4:22

Galaxea G0.5: One Autoregressive VLA for Reasoning and Action

@robotsdigest · 22 views · 2026-08-13
VLAGalaxeaG0.5
Robot-centric pointmaps align VLA perception with the robot frame
1:25

Robot-centric pointmaps align VLA perception with the robot frame

@shenzhenfoundry · 9 views · 2026-08-12
VLA具身智能点图
τ0-VLA: Robot VLA that searches over subtask sequences at test time with a world model
3:33

τ0-VLA: Robot VLA that searches over subtask sequences at test time with a world model

@jiqizhixin · 54 views · 2026-08-07
VLA世界模型World Model
GEN-1: embodied foundation model for diverse end effectors
2:09

GEN-1: embodied foundation model for diverse end effectors

@GeneralistAI · 11 views · 2026-07-23
具身智能embodied AIGEN-1