mlsanityData to 14 Oct 2026

AlphaNIST: Emergent Assembly Skills from Binary Task Rewards

arXiv 2610.13354 ↗First posted 9 Oct 2026cs.AI cs.ROcited by 0

Not on the map. Why?

Abstract (arXiv)

The NIST Assembly Task Board #1 remains a challenging benchmark for learning-based methods in robotic manipulation. Over short horizons, it demands precise, contact-rich control of small, tight-fitting parts. Over long horizons, it requires recovery and coordination across extended action sequences. Demonstrations, task decomposition, and dense progress rewards have enabled research advances by providing guidance on how to assemble. But how much of the strategy must be specified? The progression from AlphaGo to AlphaGo Zero showed that RL at scale can discover superhuman strategies without prescribing how to win. AlphaNIST asks whether a sparse binary task reward and task diversity admitted by this task-agnostic reward can similarly let a robot discover how to assemble. In simulation, we train a single policy from scratch across all 20 assembly tasks on the NIST Assembly Task Board #1, reaching 98.2% overall training success without demonstrations, progress shaping, scripted primitives, or per-task tuning. The framework also trains policies to complete ten assembly pairs in a single episode without a prescribed order, with task success rewarded only when all ten pairs are simultaneously complete. The learned policies discover dexterous and planning-like behaviors: choosing grasps for later assembly, using surrounding geometry to reorient parts, and preparing one part, assembling another, then returning to finish the first without an explicit planner. For videos, see our website: https://alpha-nist.github.io/

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 21 of its 28 references that are in the map. First posted on arXiv 9 Oct 2026.

Builds on

21 of 28 references.

References · 21

  1. 2015Deep Residual Learning for Image Recognitioncited by 52,096
  2. 2020Gradient Surgery for Multi-Task Learningcited by 1,256
  3. 2018Learning Dexterous In-Hand Manipulationcited by 687
  4. 2017Reverse Curriculum Generation for Reinforcement Learningcited by 198
  5. 2022DeXtreme: Transfer of Agile In-hand Manipulation from Simulation to Realitycited by 149
  6. 2018Multi-task Deep Reinforcement Learning with PopArtcited by 141
  7. 2023General In-Hand Object Rotation with Vision and Touchcited by 99
  8. 2024SimBa: Simplicity Bias for Scaling Up Parameters in Deep Reinforcement Learningcited by 76
  9. 2022In Defense of the Unitary Scalarization for Deep Multi-Task Learningcited by 67
  10. 2019Deep Reinforcement Learning for Industrial Insertion Tasks with Visual Inputs and Natural Rewardscited by 64
  11. 2022Factory: Fast Contact for Robotic Assemblycited by 54
  12. 2023IndustReal: Transferring Contact-Rich Assembly Tasks from Simulation to Realitycited by 53
9 more
  1. 2024DextrAH-RGB: Visuomotor Policies to Grasp Anything with Dexterous Handscited by 45
  2. 2021Robust Multi-Modal Policies for Industrial Assembly via Reinforcement Learning and Demonstrations: A Large-Scale Studycited by 30
  3. 2024AutoMate: Specialist and Generalist Assembly Policies over Diverse Geometriescited by 26
  4. 2026mjlab: A Lightweight Framework for GPU-Accelerated Robot Learningcited by 26
  5. 2024ARCH: Hierarchical Hybrid Learning for Long-Horizon Contact-Rich Robotic Assemblycited by 10
  6. 2026Emergent Dexterity via Diverse Resets and Large-Scale Reinforcement Learningcited by 10
  7. 2021Benchmarking Off-The-Shelf Solutions to Robotic Assembly Taskscited by 7
  8. 2023Towards Generalized Robot Assembly through Compliance-Enabled Contact Formationscited by 3
  9. 2025Refinery: Active Fine-tuning and Deployment-time Optimization for Contact-Rich Policiescited by 1

Built on by

Not cited yet.