Adnan KhanAdnanProduct & AI
San Francisco Bay Area
Get in touch

Source note2 min read

Learning to Fold: prizewinning solution at LeHome Challenge 2026 (1st place online, 2nd offline)

Source

arXiv

arxiv.org/abs/2606.27163v1

Read the source

Source

Plain-Language Brief

A research paper describes a system that won first place online and second place offline in the LeHome Challenge 2026, a competition on bimanual garment folding. The system improves a vision-language-action (VLA) policy using reinforcement learning.

What the Document Says

  • The paper presents a solution to the LeHome Challenge 2026, an ICRA 2026 competition focused on bimanual garment folding. The system placed first in the online round (simulation) and second in the real-world final.
  • The system improves a vision-language-action (VLA) policy with reinforcement learning.
  • The VLA policy uses the same network to predict actions and success, progress, and other task-relevant future quantities. These predictions are used for advantage estimation, live failure detection, and candidate selection.
  • The work combines existing reinforcement learning ideas with engineering and optimization contributions that can be used together or separately.
  • The paper mentions specific techniques: AWR + RECAP for flow-matching VLA, asynchronous distributed training/rollout pipeline through HuggingFace Hub, inference-time hyperparameters optimization via Thompson sampling, a sim-to-real recipe with camera-alignment tooling, heavy augmentation, and DAgger-like HIL data collection.

Key Concepts Mentioned

  • LeHome Challenge 2026
  • ICRA 2026
  • bimanual garment folding
  • Vision-Language-Action (VLA) policy
  • reinforcement learning

Claims Made by the Document

  • The system placed first in the online round and second in the real-world final of the LeHome Challenge 2026.
  • The VLA policy is improved with reinforcement learning.

Evidence Used

  • The system placed 1st of 62 teams in the online (simulation) round and 2nd in the real-world final.

Methods / Approach

UNKNOWN

Limitations Stated by the Authors

UNKNOWN

Limitations Noted by the Agent

Agent interpretation — not from the source.

UNKNOWN

Open Questions

UNKNOWN

Notes for Future Q&A

May include interpretation. Labeled where it does.

UNKNOWN

Raw Abstract / Excerpt

I describe my solution to the LeHome Challenge 2026, an ICRA 2026 competition on bimanual garment folding. The system placed 1st of 62 teams in the online (simulation) round and 2nd in the real-world final. It improves a vision-language-action (VLA) policy with a reinforcement-learning loop. The policy is its own value function: the same network that predicts actions also predicts success, progress, and a few task-relevant future quantities, and those predictions drive advantage estimation, live failure detection, and candidate selection. The work mostly recombines existing RL ideas with engineering and optimization contributions that can be used together as one recipe or individually: AWR + RECAP combined for flow-matching VLA; an asynchronous distributed training / rollout pipeline through HuggingFace Hub; inference-time hyperparameters optimization via Thompson sampling; a sim-to-real recipe with camera-alignment tooling, heavy augmentation and DAgger-like HIL data collection.