Preprint

Qantara: Bridge-Flow Training for Multi-Paradigm JEPA Control

Ruslan Rakhimov George Bredis Yuriy Maksyuta Daniil Gavrilov

Jul 2026 · 2 citations · 43 references

Computer Science

TL;DR

Qantara is presented, an end-to-end JEPA whose joint training objective pairs a Brownian-bridge interpolant between consecutive clean latents on the state axis with noise-to-data flow matching on the action axis, and these results move JEPA world models from single-paradigm planners to multi-paradigm controllers.

Abstract

Joint-Embedding Predictive Architectures (JEPAs) underpin a growing family of latent world models for control from raw pixels, but every existing JEPA world model commits at training time to a single inference paradigm: either trajectory optimisation in a learned dynamics model, or direct behaviour cloning. A single checkpoint that serves both would defer this choice to inference, when deployment constraints (rollout cost, observation accessibility) determine which path wins. We present Qantara, an end-to-end JEPA whose joint training objective pairs a Brownian-bridge interpolant between consecutive clean latents on the state axis with noise-to-data flow matching on the action axis. The same checkpoint serves three inference paradigms without retraining: latent planning, behaviour-cloning action sampling, and inverse dynamics, which we query through a video-inverse composition that first predicts the next latent without action conditioning, then extracts the action. Training concentrates mass on the edges of the (action-time, state-time) noise square, where inference queries the predictor: replacing it with uniform interior sampling drops Push-T planning from 90.1 to 53.3 SR at matched compute. On the LeWM control suite, Qantara reaches a 91.2 SR three-train-seed average and sets new SOTA on OGBench-Cube (+7.7 SR over DINO-WM, +19.7 over LeWM). From the same weights, the behaviour-cloning and video-inverse paths reach 82-83 SR on Push-T and 71-73 SR on Cube. These results move JEPA world models from single-paradigm planners to multi-paradigm controllers.

arXiv Semantic Scholar

View source

Similar papers

#artificial intelligence Preprint Aug 2026

No Gaussian Required: Contrastive Inverse Dynamics for JEPA World Models

Contrastive inverse dynamics thus provides a distribution-free anti-collapse signal that requires no target network, stop-gradient, pretrained encoder, or reconstruction objective, and it is argued that the anti-collapse pressure can instead come from the transition data itself.

Jack Boylan, Chris Hokamp · 0 citations

Preprint Jul 2026

Branch-JEPA: Finite-Support Predictive Distributions for JEPA World Models

Branch-JEPA is introduced, which replaces this point-valued transition with a context-weighted finite set of latent successors, and preserves more distinct futures, while full-set scoring improves the quality of the resulting predictive distribution.

Zhi Song, Ximing Xing, Zhenchao Tang et al. · 0 citations

Preprint Aug 2026

Diagnosing JEPA World Models with Action-Conditioned Predictive Consistency

Action-Conditioned Predictive Consistency (ACPC), a diagnostic that measures how far a clean history and a visually perturbed view of it diverge after being rolled forward under the same action sequence, is introduced and it is proved that this divergence bounds the perturbation-induced change in multi-step prediction error and planner cost.

Guo An, Zijing Wu, Hongzhuang Dong et al. · 1 citation

Preprint Aug 2026

LeFlow: Generative Latent Flow Planning for World Models

Latent world models are inherently strong encoders that transform image pixel to latent embedding, yet existing world models still rely on online trajectory optimization for action planning: for every state-goal pair, an iterative optimizer is run from scratch to search for optimal action sequences, treating the world model as a black-box simulator. This approach pays the full iterative optimization cost anew at every replanning step and reuses no planning experience across queries. In this work, we ask whether planning itself can be amortized once a latent world model has been learned. We present LeFlow, which learns a reusable latent trajectory prior operating directly in the latent dynamics space from the world model. LeFlow recasts planning as conditional latent trajectory generation: a rectified-flow model imagines a future latent path between the current and goal embeddings, an inverse dynamics decoder turns latent transitions into action chunks, and the frozen world model verifies each candidate by autoregressive rollout. Across four major goal-conditioned pixel-control benchmarks, LeFlow replaces iterative action-space optimization with amortized latent planning and fixed-budget rollout selection, achieving consistent success-rate gains with an order-of-magnitude reduction in planning time. Our results argue that latent world models should support not only prediction but reusable planning priors. Our code is available at https://github.com/hsiangwei0903/LeFlow.

Hsiang-Wei Huang, Jianxu Shangguan, Junbin Lu et al. · 0 citations

Preprint Aug 2026

JEPA-WAM: Learning Vision-Language-Action Policies with Joint-Embedding World Modeling

JEPA-WAM, a latent WAM built in a pretrained V-JEPA space, which couples latent transition prediction with continuous action generation through a shared predictor, predicts a spatially structured joint current-future target that captures task-shared visual temporal structure between current and future observations, while preserving dense patch-level correspondence.

Yihan Lin, Jiawei He, Shifeng Bao et al. · 2 citations

Jul 2026

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

InternVLA-A1.5 is presented, 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, and achieves the best overall results on all six simulation benchmarks.

Haoxiang Ma, Junhao Cai, Xiaoxu Xu et al. · arXiv.org · 10 citations · ⚡1