mlsanityData to 14 Oct 2026

Runtime-Incremental Transformer for Reinforcement-Learning-Based Adaptive Control

arXiv 2609.13545 ↗First posted 11 Sep 2026cs.RO cs.LGcited by 0

Not on the map. Why?

Abstract (arXiv)

Learning-based adaptive control of robotic manipulators with non-observable friction memory has been addressed by attention- based meta-controllers whose number of attention heads is fixed before training and is tuned by costly offline search. At long memory horizons, such fixed-capacity controllers are prone to catastrophic failures on a sizeable fraction of training seeds. The present paper introduces a runtime mechanism that grows and prunes the heads of the attention block during reinforcement learning, governed by two signals: the effective rank of the on-policy context distribution, which triggers growth when representational capacity becomes insufficient, and the per-head output magnitude, which flags redundant heads for removal. Policy continuity at growth events and a quantitative bound at prune events are established analytically. On a two- link manipulator with Stribeck friction, the proposed mechanism attains full success across all memory regimes, eliminating the long-horizon failure mode and removing the need for offline tuning of the head count.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 14 of its 31 references that are in the map. First posted on arXiv 11 Sep 2026.

Builds on

14 of 31 references.

References · 14

  1. 2018Soft Actor-Critic: Off-Policy Maximum Entropy Deep Reinforcement Learning with a Stochastic Actorcited by 4,885
  2. 2014Learning Phrase Representations using RNN Encoder-Decoder for Statistical Machine Translationcited by 4,057
  3. 2017Progressive Growing of GANs for Improved Quality, Stability, and Variationcited by 2,988
  4. 2015Learning both Weights and Connections for Efficient Neural Networkscited by 2,031
  5. 2018DARTS: Differentiable Architecture Searchcited by 1,798
  6. 2021Decision Transformer: Reinforcement Learning via Sequence Modelingcited by 1,546
  7. 2017Hindsight Experience Replaycited by 1,136
  8. 2018Efficient Neural Architecture Search via Parameter Sharingcited by 837
  9. 2021Offline Reinforcement Learning as One Big Sequence Modeling Problemcited by 644
  10. 2017Automated Curriculum Learning for Neural Networkscited by 257
  11. 2015Net2Net: Accelerating Learning via Knowledge Transfercited by 250
  12. 2021bert2BERT: Towards Reusable Pretrained Language Modelscited by 70
2 more
  1. 2026INCRT: An Incremental Transformer That Determines Its Own Architecturecited by 4
  2. 2026Temporal Attention for Adaptive Control of Euler-Lagrange Systems with Unobservable Memorycited by 1

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026Planning for Change: Reinforcement Learning Combined with Bounded Extremum Seeking for Robotic Control under Distribution Shift
  2. 2025DASIP: Dynamic Test-Time Compute Scaling for Robot Control with Stochastic Interpolant Policies
  3. 2024Deep Incremental Model Informed Reinforcement Learning for Continuous Robotic Control
  4. 2026A Balanced Data Diet: Addressing the Exploration Bottleneck in Mega-Scale RL for Robot Control
  5. 2025TARC: Time-Adaptive Robotic Control
  6. 2024Continual Policy Distillation of Reinforcement Learning-based Controllers for Soft Robotic In-Hand Manipulation
  7. 2025MEMBOT: Memory-Based Robot in Intermittent POMDP
  8. 2025XPG-RL: Reinforcement Learning with Explainable Priority Guidance for Efficiency-Boosted Mechanical Search