PEARS: Physical-Prior-Guided Efficient Adaptation via Failure Reasoning and Diffusion Steering for Tactile Manipulation

Kun Song; Yiming Wang; Yilin Chen; Tianyi Ding; Jiaxin Tian; Tianqi Gong; Daolin Ma; Jia Pan

cs.ROOct 6, 2026

PEARS: Physical-Prior-Guided Efficient Adaptation via Failure Reasoning and Diffusion Steering for Tactile Manipulation

Authors: Kun Song, Yiming Wang, Yilin Chen, Tianyi Ding, Jiaxin Tian, Tianqi Gong, Daolin Ma, Jia Pan

Organizations: The University of Hong Kong · Shanghai Jiao Tong University · Xense Robotics · University of Pennsylvania

Abstract

Pretrained robotic policies can suffer substantial performance degradation under out-of-distribution (OOD) conditions encountered during deployment, motivating post-training through real-world interaction. However, reinforcement-learning (RL)-based post-training typically requires substantial environment interactions, a burden that is especially significant in manipulation, where each trial can be slow, costly, or destructive. Therefore, we present PEARS, a physics-prior-guided hybrid RL framework for sample-efficient online adaptation of pretrained policies with tactile feedback. After each episode, its physics-guided force reasoning (PFR) module uses physical priors encoded in a vision-language model (VLM) to diagnose failures from the visual outcome and tactile interaction history and update task-appropriate contact-force bounds. A high-frequency hybrid force-position controller then enforces these bounds during contact. Complementarily, tactile-conditioned diffusion steering reinforcement learning adjusts the latent noise of the frozen flow-matching policy to correct errors in free-space motion and contact timing without updating the base model. In simulation, PEARS improves success rates by 12.4-37.4 percentage points over the strongest per-task baselines. PEARS also reduces the number of interaction episodes required for a certain success threshold by up to 53.2% relative to the fastest baseline. In real-world experiments, PEARS achieves success rates of 95% on Whiteboard Erasing and 90% on Pipette Liquid Aspiration. These results show that combining the PFR module with policy steering can accelerate adaptation while reducing costly interactions. The project website is available at https://song-kun.github.io/pears.

Figures & tables

Fig. 2: Stage-aware force control during whiteboard wiping.

Method	Thin Sheet Transfer				Bottle Cap Twisting				Fragile Fruit Picking
Method	SR (%) Mean $\pm$ Std. $\uparrow$	nAUC (%) $\uparrow$	$E_{80}$ $\downarrow$	Force failure (%) $\downarrow$	SR (%) Mean $\pm$ Std. $\uparrow$	nAUC (%) $\uparrow$	$E_{80}$ $\downarrow$	Force failure (%) $\downarrow$	SR (%) Mean $\pm$ Std. $\uparrow$	nAUC (%) $\uparrow$	$E_{80}$ $\downarrow$	Force failure (%) $\downarrow$
Base Policy	$73.7\pm 0.6$	$73.7$	/	$5.0$	$43.0\pm 1.0$	$43.0$	/	$46.0$	$23.7\pm 1.2$	23.7	/	$52.3$
Residual RL [ 6 ]	$75.7\pm 0.6$	$76.0$	12	$6.7$	$68.0\pm 20.3$	$68.0$	$47$	$20.3$	$37.3\pm 2.1$	$23.2$	/	$44.7$
AWR [ 33 ]	$67.7\pm 2.5$	$68.2$	$37$	$7.7$	$21.0\pm 13.1$	$20.3$	/	$61.7$	$15.7\pm 1.2$	$16.4$	/	$55.7$
DPPO [ 34 ]	$4.7\pm 3.5$	$2.5$	/	$5.7$	$32.7\pm 1.5$	$31.7$	/	$55.3$	$22.3\pm 2.3$	$16.3$	/	$48.0$
DSRL [ 7 ]	$82.3\pm 4.5$	$72.2$	$29$	$9.7$	$57.7\pm 1.5$	$57.9$	/	$31.3$	$40.3\pm 4.2$	$25.3$	/	$36.3$

TABLE I: Online adaptation and final-policy performance comparison. “/” denotes an unreached

E_{80}

Fig. 3: Online adaptation success across three simulation tasks. Curves show mean trailing-10 success rates across three 100-episode runs, using complete windows ending at episodes 10–100. Shaded bands denote mean

\pm

one sample standard deviation (SD)

Fig. 4: Example of force and torque intervals predicted by PFR during online interaction. a) Thin Sheet Transfer. b) Bottle Cap Twisting.

SR (%)	DSRL only	w/o DSRL	Fixed Range	Rand Search	Qwen3	PEARS
Sheet	$82.3\,{\scriptstyle\pm 4.5}$	$77.0\,{\scriptstyle\pm 1.0}$	$84.7\,{\scriptstyle\pm 3.8}$	$88.3\,{\scriptstyle\pm 6.0}$	$93.0\,{\scriptstyle\pm 1.7}$	$\mathbf{94.7}\,{\scriptstyle\pm 1.5}$
Cap	$57.7\,{\scriptstyle\pm 1.5}$	$84.7\,{\scriptstyle\pm 1.5}$	$62.3\,{\scriptstyle\pm 2.5}$	$77.0\,{\scriptstyle\pm 5.6}$	$86.7\,{\scriptstyle\pm 0.6}$	$\mathbf{87.7}\,{\scriptstyle\pm 1.5}$
Fruit	$40.3\,{\scriptstyle\pm 4.2}$	$61.7\,{\scriptstyle\pm 2.1}$	$44.0\,{\scriptstyle\pm 2.6}$	$40.3\,{\scriptstyle\pm 2.5}$	$67.0\,{\scriptstyle\pm 2.6}$	$\mathbf{77.7}\,{\scriptstyle\pm 1.2}$

TABLE II: Ablation of PFR, DSRL, force-range selection, and the VLM backbone used in PFR across three simulation tasks.

Task	Base Policy	DSRL [ 7 ]	PEARS
Whiteboard Erasing	11 / 20	15 / 20	19 / 20
Pipette Liquid Aspiration	8 / 20	12 / 20	18 / 20

TABLE III: Real-world performance (successful trials / total trials).

Explore similar work

Jul 4, 2026cs.RO

OmniTacTune: Policy-Agnostic Real-World RL for Tactile Residual Adaptation of Visual Policies

Visual policies learned from human videos, teleoperation, and robot demonstrations offer scalable motion priors, but often fail in contact-rich manipulation, where success significantly depends on local force and contact geometry. Tactile sensing provides these complementary signals, yet tactile data remain costly to collect and hard to generalize across sensors, robots, and tasks. We introduce OmniTacTune, a policy-agnostic real-world RL pipeline that adapts tactile feedback to pretrained visual policies through residual correction. OmniTacTune uses a two-stage design: it first bootstraps tactile-aware learning from autonomous base-policy rollouts, then learns a lightweight tactile residual policy through online interaction. Extensive experiments show that OmniTacTune generalizes across diverse contact-rich tasks, visual base policies, and tactile representations. Across four real-world contact-rich tasks, it improves visual base policies from 5-40% success to 85-100% within 40-80 minutes, demonstrating an efficient path for adapting tactile feedback to scalable visual robot policies. Project page: https://colinyu1.github.io/omnitactune-site/

Kelin Yu, Haode Zhang, Harish Ravichandar +2

University of Maryland, College Park · Georgia Institute of Technology

Tactile Residual Policy Teleoperation

Ask PDF

May 19, 2026cs.RO

Beyond Action Residuals: Real-World Robot Policy Steering via Bottleneck Latent Reinforcement Learning

Pretrained imitation policies have become a strong foundation for robot manipulation, but they often require online improvement to overcome execution errors, limited dataset coverage, and deployment mismatch. A central question is therefore how reinforcement learning (RL) should adapt policies after offline pretraining. Existing lightweight methods commonly apply residual corrections directly in action space, but this often leads to noisy and poorly structured exploration. In this work, we propose Z-Perturbation Reinforcement Learning (ZPRL), an approach that steers pretrained policies through a compact bottleneck latent rather than through policy weights or output actions. During offline training, we augment the policy with a plug-and-play variational information bottleneck (VIB) module to extract a task-relevant latent interface from observation embeddings. During online finetuning, the base policy is frozen and RL learns only a residual perturbation on this latent, whose decoded representation conditions the frozen action generator. We instantiate ZPRL on flow-matching policies and evaluate it on eight simulation tasks and four real-world tasks. Across diverse manipulation settings, ZPRL improves both sample efficiency and final performance over strong post-training baselines. In the real world, ZPRL improves the average success rate on four tasks by 33.7% over imitation base policies while producing smoother exploration behaviors than an action residual counterpart. These results suggest that a compact, task-aligned bottleneck latent provides an effective interface for online RL adaptation. More videos can be found at https://manutdmoon.github.io/ZPRL/.

Dongjie Yu, Kun Lei, Zhennan Jiang +2

School of Computing and Data Science, The University of Hong Kong, Hong Kong SAR. · Shanghai Qizhi Institute, Shanghai, China. · Shanghai Jiao Tong University, Shanghai, China. +2

Robot Policies Temporally Coherent Imitation Learning Robotic Manipulation Robot Systems Action Generation

Ask PDF

Sep 8, 2026cs.RO

Proxy Policy Steering

Generalist robot policies carry broad manipulation priors from large-scale data, but specializing them to a new task remains the deployment bottleneck. This requires eliciting task-specific behavior from limited demonstrations without degrading their broad capabilities. We introduce Proxy Policy Steering (PPS), an inference-time adaptation method that resolves this challenge by training two lightweight proxy policies whose calibrated velocity-space difference steers the frozen base sampler. A reference proxy models the frozen base's behavior on target-task observations, and a task proxy, initialized from the reference, captures how this behavior changes under task supervision. Their difference forms a calibrated velocity-space residual that steers the frozen base sampler at every denoising step. We identify the conditions under which this residual isolates the change induced by task supervision, and validate them empirically. Because the base is never directly modified, its broad capabilities remain available at inference, including behaviors such as recovery from failure that the demonstrations themselves do not exercise. Adaptation requires only forward velocity predictions from the base, making PPS lightweight to train and applicable even without access to the base's parameters. On 8 real-world and 4 simulation manipulation tasks, PPS lifts the state-of-the-art pi 0.5 base policy by 53% absolute success rate on average, with zero-to-one gains on tasks the base never solves, while preserving the base's broad capabilities. PPS outperforms LoRA fine-tuning, from-scratch specialists, residual policies, and prior inference-time steering methods.

Chuanruo Ning, Tianrui Wang, Wei-Chiu Ma +1

Cornell University

Robot Policies Robot Systems Steering Inference-Time

Ask PDF