# Coordinate-Based Dual-Constrained Autoregressive Motion Generation

Source: [Coordinate-Based Dual-Constrained Autoregressive Motion Generation](<https://www.sota2.com/research/paper/69d8842c7d0172140ec77aed>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-04-11T05:00:04.811Z

Authors: Kang Ding, Hongsong Wang, Jie Gui, Liang Wang

Published: 2026-04-09T11:05:29.000Z

## Abstract

Text-to-motion generation has attracted increasing attention in the research community recently, with potential applications in animation, virtual reality, robotics, and human-computer interaction. Diffusion and autoregressive models are two popular and parallel research directions for text-to-motion generation. However, diffusion models often suffer from error amplification during noise prediction, while autoregressive models exhibit mode collapse due to motion discretization. To address these limitations, we propose a flexible, high-fidelity, and semantically faithful text-to-motion framework, named Coordinate-based Dual-constrained Autoregressive Motion Generation (CDAMD). With motion coordinates as input, CDAMD follows the autoregressive paradigm and leverages diffusion-inspired multi-layer perceptrons to enhance the fidelity of predicted motions. Furthermore, a Dual-Constrained Causal Mask is introduced to guide autoregressive generation, where motion tokens act as priors and are concatenated with textual encodings. Since there is limited work on coordinate-based motion synthesis, we establish new benchmarks for both text-to-motion generation and motion editing. Experimental results demonstrate that our approach achieves state-of-the-art performance in terms of both fidelity and semantic consistency on these benchmarks.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Temporal Motion Inpainting on HumanML3D](<https://www.sota2.com/research/sota/temporal-motion-inpainting-on-humanml3d.md>): Temporal Motion Inpainting; HumanML3D; FID: 0.103.

- [Temporal Motion Outpainting on HumanML3D](<https://www.sota2.com/research/sota/temporal-motion-outpainting-on-humanml3d.md>): Temporal Motion Outpainting; HumanML3D; FID: 0.104.

- [Temporal Motion Suffix on HumanML3D](<https://www.sota2.com/research/sota/temporal-motion-suffix-on-humanml3d.md>): Temporal Motion Suffix; HumanML3D; FID: 0.09.

- [Temporal Motion Prefix on HumanML3D](<https://www.sota2.com/research/sota/temporal-motion-prefix-on-humanml3d.md>): Temporal Motion Prefix; HumanML3D; FID: 0.163.

- [Text-to-motion generation on KIT (test)](<https://www.sota2.com/research/sota/text-to-motion-generation-on-kit-test.md>): Text-to-motion generation; KIT (test); R-Precision Top-1: 41.6.

- [Text-to-motion generation on HumanML3D (test)](<https://www.sota2.com/research/sota/text-to-motion-generation-on-humanml3d-test.md>): Text-to-motion generation; HumanML3D (test); FID: 0.046.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2604.08088>)

[PDF](<https://arxiv.org/pdf/2604.08088.pdf>)
