AI Paper Atlas

Research map / Reinforcement learning / Visual and representation RL

RoboCasa365: A Large-Scale Simulation Framework for Training and Benchmarking Generalist Robots

Soroush Nasiriany, Sepehr Nasiriany, Abhiram Maddukuri, Yuke Zhu

ICLR 2026 · Reinforcement learning › Visual and representation RL · approach: Dexterous manipulation transfer · 61 citations (OpenAlex)

Read the paper

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use GR00T N1.5 instead of Diffusion Policymulti-task policy · 300-task multi-task averagesuccess rate: 6.1 -> 20.0Improved
pretrain on RoboCasa365 data before target fine-tuning (100% target) (ablation)GR00T N1.5 · target averagesuccess rate: 43.7 -> 51.1Improved
pretrain before fine-tuning with 10% target data (ablation)GR00T N1.5 · target averagesuccess rate: 21.0 -> 35.9Improved
add MimicGen synthetic data to human pretraining (10% target) (ablation)GR00T N1.5 · target averagesuccess rate: 40.0 -> 35.9Hurt
add MimicGen synthetic data to human pretraining (100% target) (ablation)GR00T N1.5 · target averagesuccess rate: 52.5 -> 51.1Hurt
increase pretraining tasks from 50 to 300 (10% target) (ablation)GR00T N1.5 · target averagesuccess rate: 34.7 -> 40.0Improved
increase pretraining scenes from 5 to 2500 (ablation)GR00T N1.5 · zero-shot target kitchenssuccess rate: 29.6 -> 44.7Improved
use full fine-tuning instead of LoRA (ablation)GR00T N1.5 · multi-task averagesuccess rate: 1.2 -> 20.0Improved
use two-stage pretrain then fine-tune instead of joint co-training (ablation)GR00T N1.5 · target averagesuccess rate: 22.5 -> 51.1Improved
add sim data to real data for real-robot trainingGR00T N1.5 · real-world 4 taskssuccess rate: 61.8 -> 79.8Improved
learn longer-horizon tasks sequentiallyGR00T N1.5 · lifelong learning atomic taskssuccess rate: 41.5 -> 10.6Hurt
perturb camera pose at test time (ablation)GR00T N1.5 · Composite-Seensuccess rate: 40.6 -> 28.8Hurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy