IVRA

Improving Visual-Token Relations for Robot Action Policy with Training-Free Hint-Based Guidance

Jongwoo Park, Kanchana Ranasinghe, Jinhyeok Jang, Cristina Mata, Yoo Sung Jang, Michael S Ryoo

arXiv Code More Demos BibTeX

Demo Videos

With IVRA (Success)
IVRA (Affinity Hint)
LLaRA Baseline (Fail)
IVRA affinity hint map for the cluttered localization task.

Cluttered Localization Task

<Pick up the yellow duck and drop it into a pan>

IVRA affinity hint map for the relative height short-object task.

Relative Height Task

<Pick up the short object and place it on the pan>

IVRA affinity hint map for the relative height long-object task.

Relative Height Task

<Pick up the long object and place it on the pan>

IVRA affinity hint map for the color match task.

Color Match Task

<Pick up object same color as the duck and drop it into a pan>

Abstract

Many Vision-Language-Action (VLA) models flatten image patches into a 1D token sequence, weakening the 2D spatial cues needed for precise manipulation. We introduce IVRA, a lightweight, training-free method that improves spatial understanding by exploiting affinity hints already available in the model's built-in vision encoder, without requiring any external encoder or retraining. IVRA selectively injects these affinity signals into a language-model layer in which instance-level features reside. This inference-time intervention realigns visual-token interactions and better preserves geometric structure while keeping all model parameters fixed. We demonstrate the generality of IVRA by applying it to diverse VLA architectures (LLaRA, OpenVLA, and FLOWER) across simulated benchmarks spanning both 2D and 3D manipulation (VIMA and LIBERO) and on various real-robot tasks. On 2D VIMA, IVRA improves average success by +4.2% over the baseline LLaRA in a low-data regime. On 3D LIBERO, it yields consistent gains over the OpenVLA and FLOWER baselines, including improvements when baseline accuracy is near saturation (96.3% to 97.1%).

Architecture

IVRA architecture overview showing affinity-guided token mixing and sharper affinity maps after applying IVRA.
A frozen vision encoder provides an affinity hint that guides token mixing with weighted pooled tokens, preserving instance-level cues and improving manipulation policy quality. Brighter regions indicate higher affinity relative to the reference point (red dot).

Additional IVRA Demos

BibTeX

@article{park2026ivra,
  title={IVRA: Improving Visual-Token Relations for Robot Action Policy with Training-Free Hint-Based Guidance},
  author={Park, Jongwoo and Ranasinghe, Kanchana and Jang, Jinhyeok and Mata, Cristina and Jang, Yoo Sung and Ryoo, Michael S},
  journal={arXiv preprint arXiv:2601.16207},
  year={2026}
}