LMM-Track4D: Eliciting 4D Dynamic Reasoning in LMMs via Trajectory-Grounded Dialogue

• URL: http://arxiv.org/abs/2605.19390v1
• Date: Tue, 19 May 2026 05:35:13 GMT
• Title: LMM-Track4D: Eliciting 4D Dynamic Reasoning in LMMs via Trajectory-Grounded Dialogue
• Abstract summary: Recent models (LMMs) have become increasingly capable on image video understanding, yet still struggle to sustain 4D continuous dynamic reasoning.<n>To study this capability, we propose a new task in which a model must answer queries while returning 3D trajectories over an entire short or a specified segment of a longer clip.<n>Experiments on Track4-Track show consistent improvements over strong baselines, suggesting that dynamic state modeling is a useful design principle for eliciting 4D dynamic reasoning in LMMs.
• Score: 4.481491966887467
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Recent large multimodal models (LMMs) have become increasingly capable on image and video understanding, yet still struggle to sustain 4D continuous spatiotemporal dynamic reasoning. To study this capability gap, we formulate trajectory-grounded multi-turn spatiotemporal dialogue, a new task in which a model must answer spatiotemporal queries while returning structured 3D target trajectories over an entire short clip or a specified segment of a longer clip, and introduce Track4D-Bench, a benchmark with 526 clip-level dialogue samples spanning 23.5k frames and 7.5k object annotations, for training and evaluation. Building on this task, we propose LMM-Track4D, which combines RTGE (Ray--Time Geometry Encoding), a dedicated streaming state token TRK for long-horizon dynamic propagation, and an Object-Slot Kinematic, Residual-Anchor (OSK-RA) decoder for stable 4-step 3D state estimation under occlusion and viewpoint variation. Experiments on Track4D-Bench show consistent improvements over strong baselines, suggesting that explicit dynamic state modeling is a useful design principle for eliciting 4D dynamic reasoning in LMMs. Our code and dataset will be publicly available at https://github.com/mikubaka88/LMM-Track4D.

Related papers

• DynTrace: Tracking Dynamic Object Evidence for 4D Spatio-Temporal Reasoning in MLLMs [41.4243802276804]
  4D-temporal reasoning is essential for understanding dynamic worlds and enabling embodied interaction.<n>Current models rely on sparse frame-level observations, fragmenting continuous dynamic cues.<n>We propose DynTrace, a training-free framework for 4D-temporal reasoning with two complementary components.
  arXiv  Detail & Related papers  (2026-07-14T08:35:07Z)
• MLLM-4D: Towards Visual-based Spatial-Temporal Intelligence [50.11889361459544]
  Humans are born with vision-based 4D spatial-temporal intelligence.<n>Despite its importance, this capability remains a significant bottleneck for current large language models (MLLMs)
  arXiv  Detail & Related papers  (2026-02-28T07:23:36Z)
• Learning to Reason in 4D: Dynamic Spatial Understanding for Vision Language Models [79.18306680174011]
  DSR Suite bridges gap across aspects of dataset, benchmark and model.<n>We propose an automated pipeline that generates multiple-choice question-answer pairs from in-the-wild videos for DSR.<n>The pipeline extracts rich geometric and motion information, including camera poses, local point clouds, object masks, orientations, and 3D trajectories.
  arXiv  Detail & Related papers  (2025-12-23T17:56:36Z)
• C4D: 4D Made from 3D through Dual Correspondences [77.04731692213663]
  We introduce C4D, a framework that leverages temporal correspondences to extend existing 3D reconstruction formulation to 4D.<n>C4D captures two types of correspondences: short-term optical flow and long-term point tracking.<n>We train a dynamic-aware point tracker that provides additional mobility information.
  arXiv  Detail & Related papers  (2025-10-16T17:59:06Z)
• St4RTrack: Simultaneous 4D Reconstruction and Tracking in the World [106.91539872943864]
  St4RTrack is a framework that simultaneously reconstructs and tracks dynamic video content in a world coordinate frame from RGB inputs.<n>We predict both pointmaps at the same moment, in the same world, capturing both static and dynamic scene geometry.<n>We establish a new extensive benchmark for world-frame reconstruction and tracking, demonstrating the effectiveness and efficiency of our unified, data-driven framework.
  arXiv  Detail & Related papers  (2025-04-17T17:55:58Z)
• Easi3R: Estimating Disentangled Motion from DUSt3R Without Training [69.51086319339662]
  We introduce Easi3R, a simple yet efficient training-free method for 4D reconstruction.<n>Our approach applies attention adaptation during inference, eliminating the need for from-scratch pre-training or network fine-tuning.<n>Our experiments on real-world dynamic videos demonstrate that our lightweight attention adaptation significantly outperforms previous state-of-the-art methods.
  arXiv  Detail & Related papers  (2025-03-31T17:59:58Z)
• Diffusion4D: Fast Spatial-temporal Consistent 4D Generation via Video Diffusion Models [116.31344506738816]
  We present a novel framework, textbfDiffusion4D, for efficient and scalable 4D content generation. We develop a 4D-aware video diffusion model capable of synthesizing orbital views of dynamic 3D assets. Our method surpasses prior state-of-the-art techniques in terms of generation efficiency and 4D geometry consistency.
  arXiv  Detail & Related papers  (2024-05-26T17:47:34Z)
• MAMBA4D: Efficient Long-Sequence Point Cloud Video Understanding with Disentangled Spatial-Temporal State Space Models [14.024240637175216]
  We propose a novel point cloud video understanding backbone based on the State Space Models (SSMs)<n> Specifically, we first disentangle space and time in 4D video sequences and then establish the spatial-temporal correlation with our designed Mamba blocks.<n>Our method has a significant efficiency improvement with 87.5% GPU memory reduction and 5.36 times speed-up.
  arXiv  Detail & Related papers  (2024-05-23T09:08:09Z)
• Consistent4D: Consistent 360{\deg} Dynamic Object Generation from Monocular Video [15.621374353364468]
  Consistent4D is a novel approach for generating 4D dynamic objects from uncalibrated monocular videos. We cast the 360-degree dynamic object reconstruction as a 4D generation problem, eliminating the need for tedious multi-view data collection and camera calibration.
  arXiv  Detail & Related papers  (2023-11-06T03:26:43Z)
• NSM4D: Neural Scene Model Based Online 4D Point Cloud Sequence Understanding [20.79861588128133]
  We introduce a generic online 4D perception paradigm called NSM4D. NSM4D serves as a plug-and-play strategy that can be adapted to existing 4D backbones. We demonstrate significant improvements on various online perception benchmarks in indoor and outdoor settings.
  arXiv  Detail & Related papers  (2023-10-12T13:42:49Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.