AI Paper Atlas

Research map / Video / Video generation

Reconstruction-Anchored Diffusion Model for Text-to-Motion Generation

Yifei Liu, Changxing Ding, Ling Guo, Huaiguang Jiang, Qiong Cao

ECCV 2026 · Video › Video generation · approach: Text-to-motion diffusion · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add motion encoder branch (ablation)text encoder + diffusion baseline · HumanML3D testFID: 0.786 -> 0.624Improved
add motion-centric latent alignment L_latent (ablation)baseline + E_m · HumanML3D testFID: 0.624 -> 0.187Improved
add self-regularization L_sr (ablation)baseline + E_m + L_latent · HumanML3D testFID: 0.187 -> 0.132Improved
add Reconstructive Error Guidance (ablation)RAM without REG · HumanML3D testFID: 0.132 -> 0.032Improved
add REG (ablation)RAM · HumanML3D testR-Precision Top-1: 0.561 -> 0.561No clear effect
add classifier-free guidance (w2=1.5) (ablation)RAM without guidance · HumanML3D testR-Precision Top-1: 0.529 -> 0.561Improved
add REG only in first 6 denoising steps (ablation)RAM (20 steps) · HumanML3D testFID: 0.132 -> 0.038Improved
add bidirectional alignment with weight 1.0 to two-stream baseline (ablation)Variant A · HumanML3D testFID: 0.624 -> 0.776Hurt
add cross-modal contrastive learning to two-stream baseline (ablation)Variant A · HumanML3D testFID: 0.624 -> 0.708Hurt
use motion-centric alignment plus self-regularization instead of contrastive+bidirectional (ablation)Variant E · HumanML3D testFID: 0.218 -> 0.132Improved
reduce beta from 1.00 to 0.01 (limit gradient into motion encoder) (ablation)RAM · HumanML3D testFID: 0.278 -> 0.032Improved
set self-regularization weight to nonzero (1.0 vs 0.0) (ablation)RAM · HumanML3D testFID: 0.109 -> 0.037Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 2 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy