TinyAstral: Transferring Self-Evolving Harnesses to Tiny Generalist Manipulation Agents

Kehong Gong; Xiaoxu Zheng; Guanli Hou; Seunghun Lee; Mengqi Guo; Runjia Tan; Zhengyu Wen; Qi Wang; Xiaoyu He; Mingyuan Zhang; Michael Bi Mi

arXiv · 2610.41216

TinyAstral: Transferring Self-Evolving Harnesses to Tiny Generalist Manipulation Agents

Kehong Gong·Xiaoxu Zheng·Guanli Hou·Seunghun Lee·Mengqi Guo·Runjia Tan·Zhengyu Wen·Qi Wang·Xiaoyu He·Mingyuan Zhang·Michael Bi Mi

View original source ↗Read source document ↗Export metadata ↓

Abstract

Robot agents built on vision-language models depend on both the planner and its harness: the tools, feedback, memory, and recovery logic around it. Can improvements to a large teacher's harness be carried into a compact agent with lower inference compute cost? We present TinyAstral, a two-stage approach. We first freeze an open-weight teacher and improve its harness through recursive self-improvement (RSI): frontier LLM agents read the teacher's trajectories and propose harness edits, and each edit is accepted only if it passes a paired statistical test. We then freeze the final harness, collect teacher trajectories with reasoning and failure recovery, and distill a small student that replaces the teacher in the same harness. On RoboTwin, harness evolution raises the teacher's success from 33.4% to 63.6%; the final 4B student reaches 62.8%. Across four harness versions, distilled 4B students score close to their corresponding teachers, and later versions yield stronger students, showing that harness evolution gains carry into compact models. The harness gains also transfer to other frozen planners without retraining. On identical episodes, the final 4B student outperforms open 4B agents, the open 27B harness-RSI model SenseNova-RoboRSI, and our own teacher model placed in three competitors' harnesses. It stays close to its teacher on all three training benchmarks, and on three benchmarks never used for RSI or training it scores close to a 552B planner running the same harness. On the measured deployment episodes, it uses about one fifth of its teacher's accelerator time.

Explore related subjects

cs.RO cs.CV

Keep this discovery

Explore connections, maps & timelines

BibTeX RIS

Kehong Gong, Xiaoxu Zheng, Guanli Hou, Seunghun Lee, Mengqi Guo, Runjia Tan, Zhengyu Wen, Qi Wang, Xiaoyu He, Mingyuan Zhang, Michael Bi Mi. 2026-10-30. TinyAstral: Transferring Self-Evolving Harnesses to Tiny Generalist Manipulation Agents. https://arxiv.org/abs/2610.41216

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related papers

arXiv2026

Design of an Underwater Vehicle With Orientation-Adjustable Thrusters and Adaptive Tracking Control

Unmanned underwater vehicles (UUVs) are essential for marine exploration and research. However, conventional designs often struggle with limited maneuverability in complex and dynamic underwater environments. This paper presents an orientation-adjustable thruster UUV (OAT-UUV) with vector thrusters, enabling full six-degree-of-freedom (6-DOF) motion. and composite maneuvers. To overcome challenges associated with uncertain model parameters and environmental disturbances, a vision-based tracking control system based on a novel feedforward adaptive model predictive controller (FF-AMPC) is proposed to achieve robust object tracking, which integrates realtime state feedback with adaptive parameter updating. Extensive experiments, including closed-loop tracking and composite motion tests in real-world environment, have been conducted. The obtained results demonstrate that the orientation-adjustable propulsion design significantly improves the underwater maneuverability, while the FF-AMPC controller achieves a 49.1% hovering improvement compared to conventional controller.

cs.RO↗

arXiv2026

FlowDrive: moderated flow matching with data balancing for trajectory planning

Learning-based planners are sensitive to the long-tailed distribution of driving data. Common maneuvers dominate datasets, while dangerous or rare scenarios are sparse. This imbalance can bias models toward the frequent cases and degrade performance on critical scenarios. To tackle this problem, we compare balancing strategies for sampling training data and find reweighting by trajectory pattern an effective approach. We then present FlowDrive, a flow-matching trajectory planner that learns a conditional rectified flow to map noise directly to trajectory distributions with few flow-matching steps. We further introduce moderated, in-the-loop guidance that injects small perturbation between flow steps to systematically increase trajectory diversity while remaining scene-consistent. On nuPlan and the interaction-focused interPlan benchmarks, FlowDrive achieves state-of-the-art results among learning-based planners and approaches methods with rule-based refinements. After adding moderated guidance and light post-processing (FlowDrive*), it achieves overall state-of-the-art performance across nearly all benchmark splits. Our code is available at https://github.com/einsteinguang/flow_drive_planner.

cs.RO↗

arXiv2026

DexCanvas: Bridging Human Demonstrations and Robot Learning for Dexterous Manipulation

A dexterous hand succeeds or fails through contact, yet every scalable record of human manipulation (video, vision-based hand tracking, motion capture) observes how the hand moves and misses how it pushes. We present DexCanvas, a hybrid real-synthetic dataset of 7,000 hours of dexterous hand-object interaction with per-frame contact points and contact forces, grown from 70 hours of optical motion capture that spans 21 manipulation types, from power and precision grasps to in-hand manipulation. The forces come from a real-to-sim pipeline: one reinforcement learning (RL) policy, conditioned on object geometry and manipulation type, drives an actuated MANO hand in physics simulation to reproduce each captured object trajectory, and the simulator measures the contacts that produce the reproduced motion. The policy reproduces 87.5% of the demonstrations and still succeeds on 80.9% of rollouts from perturbed initial object poses; these perturbed rollouts expand the 70 captured hours a hundredfold. On six YCB objects absent from the capture set it succeeds 72.5% of the time. The contact channels make better policies: a Diffusion Policy that observes DexCanvas contact features succeeds 82.2% of the time, against 67.5% without them, over 18,522 trials. Contact-driven retargeting carries the same demonstrations to two dexterous robot hands, where the RL stage again produces stable rollouts. A first release of 111,857 force-annotated trajectories is publicly available on Hugging Face.

cs.RO↗