EATR-Stereo: Embodiment-Aware Token Routing of Paired Stereo Evidence for Humanoid Vision-Language-Action Control

Songwei Wu; Rui Zhao; Fan Yang; Zhongqiang Nie; Zhiduo Jiang; Wandong Sun; Yuwei Li; Jian Hu

cs.ROAug 18, 2026

EATR-Stereo: Embodiment-Aware Token Routing of Paired Stereo Evidence for Humanoid Vision-Language-Action Control

Authors: Songwei Wu, Rui Zhao, Fan Yang, Zhongqiang Nie, Zhiduo Jiang, Wandong Sun, Yuwei Li, Jian Hu,

Organizations: Harbin Institute of Technology · Honor

Abstract

Long-horizon humanoid vision--language--action (VLA) control with head-mounted stereo cameras requires visual interfaces that can exploit complementary views while maintaining compatibility with pretrained representations. Existing interfaces often discard complementary stereo evidence or fuse additional observations without preserving the native primary-view pathway and adapting auxiliary information to robot embodiment. We present EATR-Stereo, an embodiment-aware token-routing framework that retains primary-view tokens and constructs primary-aligned Cross-View Auxiliary Tokens (CVATs) by querying the synchronized auxiliary-view token sequence. A body-segmented proprioceptive encoder further conditions token-wise auxiliary usage on robot configuration history, enabling selective incorporation of stereo evidence during action generation. The routed auxiliary stream augments the language and primary-visual context of a pretrained VLA while keeping its vision--language model frozen. On a 33-DoF physical humanoid with a 37-D proprioceptive state, we evaluate nine configurations in over-100-s search--approach--grasp--place--return tasks. EATR-Stereo achieves 60.0% full-task success, 100.0% grasp success, and 80.0% stage success. Under severe asymmetric occlusion, it improves recovery to 80% compared with 30% for CVAT alone. Ablation studies further show the importance of preserving primary tokens and combining cross-view auxiliary features with structured proprioceptive routing. These results demonstrate that selectively routed paired stereo evidence improves spatial grounding for reliable long-horizon humanoid VLA control.

Figures & tables

Fig. 1: Motivation for paired stereo in humanoid VLA control. (A) Stationary-camera arm settings do not capture the head-camera dynamics of bipedal whole-body motion. (B) Whole-body motion makes head-mounted visual evidence configuration-dependent and consequential over long horizons. (C) Synchronized stereo supplies paired geometric and visibility cues under these dynamics. These conditions motivate an interface that relates synchronized observations, preserves primary-view semantics, and applies body-segmented proprioceptive routing to auxiliary evidence.

Fig. 2: Overview of EATR-Stereo and its deployment pipeline. (A) A frozen pretrained VLM encodes the synchronized primary and auxiliary views with the language instruction. Primary-query cross-view attention constructs a separate, primary-aligned CVAT stream while preserving the native primary tokens. Body-segmented proprioceptive routing conditions token-wise CVAT usage before a shared flow-matching action head predicts an action chunk. (B) At deployment, an external VLM evaluates subtask completion and switches the instruction through a slow-timescale loop, while asynchronous buffering and local B-Spline stitching support fast-timescale control. These deployment components lie outside learned stereo routing.

Fig. 3: Real-humanoid long-horizon evaluation under a safety-assisted protocol allowing at most two interventions per trial. (A) The HONOR Omega 1.0 platform with synchronized head-mounted stereo cameras and BeyondMimic whole-body control. (B) The search–approach–grasp–place–return task, which lasts more than 100 s. (C) Demonstration-covered (ID) and unseen (OOD) bottle positions tested in a fixed order; annotations for EATR-Stereo report 10/10 grasp success in each split and 80% basket-color compliance. (D) Targeted severe asymmetric-occlusion analysis, where the bottle is fully hidden in the primary view but remains partially visible in the auxiliary view; EATR-Stereo recovers in 8/10 trials, compared with 3/10 for CVAT without state routing.

Method	Success (%) $\uparrow$	Success/Trials
EATR-Stereo	43.33	156/360
CVAT	39.44	142/360
StereoPolicy	38.06	137/360
CVAT-Flat	37.50	135/360
QwenPI0.5	36.11	130/360
GR00T1.7	35.56	128/360

TABLE I: Top-seven RoboCasa365 results over all 18 tasks.

Method	Full-task Success $\uparrow$	Grasp Success $\uparrow$	Stage Success Rate $\uparrow$	ID Stage Success $\uparrow$	OOD Stage Success $\uparrow$
GR00T1.7-Mono	35.0%	55.0%	45.0%	55.0%	35.0%
GR00T-Wide	20.0%	45.0%	32.5%	40.0%	25.0%
GR00T	35.0%	80.0%	57.5%	75.0%	40.0%
StereoPolicy	45.0%	85.0%	65.0%	80.0%	50.0%
EATR-Stereo	60.0%	100.0%	80.0%	90.0%	70.0%

TABLE II: Main physical-humanoid comparison over 20 trials per method. ID and OOD denote demonstration-covered and unseen bottle positions, respectively.

Fig. 4: Physical-humanoid ablations over 20 trials per method. CVAT-F denotes CVAT-Flat and Stereo-R denotes Stereo-Route.

	Position Generalization			Semantic Compliance
Method	ID Grasp (%) $\uparrow$	OOD Grasp (%) $\uparrow$	Gap (pp) $\downarrow$	Left/Right (%) $\uparrow$	Color (%) $\uparrow$
GR00T1.7-Mono	70	40	$30$	100	80
GR00T-Wide	50	40	$10$	0	0
GR00T	100	60	$40$	100	80
StereoPolicy	100	70	$30$	60	0
Main-XAttn	60	60	$0$	0	0

TABLE III: Position generalization and semantic compliance on the physical humanoid. Grasp success is measured over 10 trials per position split. Semantic compliance is measured over 10 trials per instruction type and denotes selection of the basket on the instructed side (Left/Right) or with the instructed color (Color). Gap denotes the ID-to-OOD drop in grasp success.

Method	Recovery Success $\uparrow$	Mean Time (s) $\downarrow$
CVAT	30% (3/10)	$>55.0$
CVAT-Flat	60% (6/10)	41.7
EATR-Stereo	80% (8/10)	22.4

TABLE IV: Severe asymmetric-occlusion recovery. Mean time uses successful trials only; all trials use a 90-s timeout.

Method	VLM Adapt.	Time (h)	Relative	Stage (%) $\uparrow$
GR00T	Frozen	10.35	$1.00\times$	57.5
EATR-Stereo	Frozen	10.95	$1.06\times$	80.0
Estimated-Depth VLA	4 layers	41.80	$4.04\times$	80.0

TABLE V: System-level training cost and physical-task performance.

Fig. 5: Qualitative action-chunk continuity for a representative joint around a chunk boundary.

Robotic Control

Robotic control studies how machines use sensing and computation to produce reliable movement and interaction in changing environments.

See all 562 papers

StereoVLA: Enhancing Vision-Language-Action Models with Stereo Vision
Dec 26, 2025Shengliang Deng, Mi Yan, Yixin Zheng +7Visual Spatial Reasoning Robotic Manipulation
Galbot · The University of Hong Kong · CFCS, School of CS, Peking University +3
ExStereo: Lifting 2D Vision-Language-Action Models to 3D with Explicit Stereo Representations
Oct 3, 2026I-Chun Arthur Liu, Jason Chen, Gaurav S. Sukhatme +1Vision-Language-Action Models Robotic Manipulation
Viterbi School of Engineering, University of Southern California
Not All Features Are Created Equal: A Mechanistic Study of Vision-Language-Action Models
Mar 19, 2026Bryce Grant, Xijia Zhao, Peng WangVision-Language-Action Models Robotics
Case Western Reserve University

Dec 26, 2025J/K move · Enter open · S save

StereoVLA: Enhancing Vision-Language-Action Models with Stereo Vision

Shengliang Deng, Mi Yan, Yixin Zheng, Jiayi Su+6

Galbot · The University of Hong Kong · CFCS, School of CS, Peking University+3

Visual Spatial Reasoning Robotic Manipulation Stereo Vision Vision-Language-Action Models VLMs for Robotics

Open paper AskPDF arXiv

While Vision-Language-Action (VLA) models excel in generalist manipulation, they often lack fine-grained spatial awareness and show limited viewpoint robustness. This limitation largely stems from the reliance on pretrained RGB encoders, which lack explicit geometric cues and prioritize semantic alignment over geometric representation. We argue that effective visual representations for VLA models must jointly encode both semantic and geometric information. In this paper, we introduce StereoVLA, the first VLA model to incorporate rich geometric cues from large-scale synthetic stereo data. StereoVLA employs a Geometric-and-Semantic (GeoSem) vision encoder that extracts geometric cues from subtle stereo-view disparities for precise spatial perception, while simultaneously capturing semantic features from pixel observations to support language-conditioned manipulation. Additionally, we introduce two synergistic co-training objectives: Interaction-Region Depth Estimation for precise spatial reasoning, and Camera Parameter Estimation to implicitly align perception and action coordinate systems. Compared with baselines that employ various input modalities, StereoVLA achieves a 33.4% absolute gain in success rate in real-world experiments and demonstrates robustness to near-hemispheric camera perspectives. Project page: https://shengliangd.github.io/StereoVLA-Webpage.

Abstract

Robotic Control

Explore similar work

StereoVLA: Enhancing Vision-Language-Action Models with Stereo Vision

ExStereo: Lifting 2D Vision-Language-Action Models to 3D with Explicit Stereo Representations

Not All Features Are Created Equal: A Mechanistic Study of Vision-Language-Action Models