InternVLA-A1.5: Unifying Understanding, Latent Foresight, and Action for Compositional Generalization

Hugging Face Daily Papers Papers

Summary

InternVLA-A1.5 integrates pretrained vision-language models with future prediction in latent space to enable efficient robot manipulation with compositional generalization and long-horizon execution, achieving state-of-the-art results on simulation benchmarks.

Unified models for robot manipulation aim to equip one policy with both the semantic priors of pretrained VLMs and the physical dynamics learned through future prediction. In practice, existing designs tend to erode the semantics of the pretrained backbone, suffer interference among heterogeneous objectives, and learn future prediction from scratch in pixel space, leaving the dynamics priors of pretrained video generators unexploited. We present InternVLA-A1.5, which builds the policy on a native VLM backbone that keeps training on VQA and subtask prediction, and attaches a lightweight unified expert for continuous action generation. Future prediction is recast as a latent-querying problem, where a small set of learnable foresight tokens condenses the task-relevant future into a compact latent code under the supervision of a frozen pretrained video generation model, so the policy inherits world-model dynamics priors without ever learning pixel-level generation. The video branch is discarded at inference, keeping real-time control. Pretrained on 1.2M robot episodes and 3M multimodal samples, InternVLA-A1.5 achieves the best overall results on all six simulation benchmarks. In the real world, the preserved semantics deliver the strongest compositional generalization on held-out instruction bindings, and the two designs together sustain long-horizon execution.
Original Article
View Cached Full Text

Cached at: 07/07/26, 06:42 AM

Paper page - InternVLA-A1.5: Unifying Understanding, Latent Foresight, and Action for Compositional Generalization

Source: https://huggingface.co/papers/2607.04988 Authors:

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

Abstract

InternVLA-A1.5 integrates pretrained vision-language models with future prediction in latent space to enable efficient robot manipulation with preserved semantics and long-horizon execution.

Unified models for robot manipulation aim to equip one policy with both the semantic priors of pretrained VLMs and the physical dynamics learned throughfuture prediction. In practice, existing designs tend to erode the semantics of the pretrained backbone, suffer interference among heterogeneous objectives, and learnfuture predictionfrom scratch in pixel space, leaving the dynamics priors of pretrained video generators unexploited. We present InternVLA-A1.5, which builds the policy on a nativeVLM backbonethat keeps training on VQA and subtask prediction, and attaches a lightweightunified expertforcontinuous action generation.Future predictionis recast as alatent-querying problem, where a small set of learnableforesight tokenscondenses the task-relevant future into a compact latent code under the supervision of a frozenpretrained video generation model, so the policy inheritsworld-model dynamics priorswithout ever learning pixel-level generation. The video branch is discarded at inference, keeping real-time control. Pretrained on 1.2Mrobot episodesand 3Mmultimodal samples, InternVLA-A1.5 achieves the best overall results on all six simulation benchmarks. In the real world, the preserved semantics deliver the strongest compositional generalization on held-out instruction bindings, and the two designs together sustain long-horizon execution.

View arXiv pageView PDFProject pageAdd to collection

Get this paper in your agent:

hf papers read 2607\.04988

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper1

#### InternRobotics/InternVLA-A1.5-base Robotics• 3B• Updatedabout 3 hours ago • 6 • 5

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2607.04988 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2607.04988 in a Space README.md to link it from this page.

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.

Similar Articles

IntentVLA: Short-Horizon Intent Modeling for Aliased Robot Manipulation

Hugging Face Daily Papers

IntentVLA is a history-conditioned visual-language-action framework that improves robot imitation learning stability by encoding short-horizon intents from visual observations, addressing challenges from partial observability and ambiguous observations. It also introduces AliasBench, an ambiguity-aware benchmark for evaluating such methods.

From Foundation to Application: Improving VLA Models in Practice

Papers with Code Trending

This paper presents LingBot-VLA 2.0, which enhances VLA foundation models for robotics by improving generalization across tasks and embodiments, expanding action space to whole-body degrees of freedom, and incorporating predictive dynamics modeling for better temporal reasoning.