--- title: '3DEditVerse: Scalable 3D Editing Benchmark' url: https://www.emergentmind.com/topics/3deditverse type: topic --- # 3DEditVerse: Scalable 3D Editing Benchmark Searching arXiv for the cited works to ground the article in recent papers. 3DEditVerse is a paired benchmark and data-generation framework for localized 3D asset editing, introduced to support scalable learning and evaluation of edits that modify geometry or appearance while preserving unedited structure and maintaining multi-view consistency. In the formulation associated with "Towards Scalable and Consistent 3D Editing" [2510.02994], 3DEditVerse provides before–after 3D correspondences rather than unrelated assets plus text, and is designed around three target properties: edit locality, multi-view consistency, and semantic alignment. In subsequent work, it also functions as supervised editing data for native 3D editors such as Omni123, which fine-tunes an autoregressive text-and-3D model on 3DEditVerse for instruction-based source-mesh-to-target-mesh editing [2604.02289]. ## 1. Definition and scope 3DEditVerse is defined as a large-scale paired 3D editing benchmark in which each sample contains a source 3D asset and a target edited 3D asset related by a localized transformation rather than independent generation [2510.02994]. The benchmark is explicitly motivated by the claim that 3D editing differs from 2D editing in requiring cross-view consistency, structural fidelity in preserved regions, fine-grained controllability, and practical scalability. Within that framing, paired supervision is treated as the critical missing ingredient: it specifies both what should change and what should remain fixed. The benchmark is described as the largest paired 3D editing benchmark to date, with **116,309 training pairs** and **1,500 curated test pairs** [2510.02994]. The same paper also reports an underlying corpus of about **118K paired 3D assets**, decomposed into **54,000** structural or pose-driven assets and **64,123** appearance-oriented edits; the benchmark split numbers remain the operative training/test counts. The benchmark spans both geometric and appearance edits, and its intended use is twofold: supervised training of 3D editors and standardized quantitative evaluation of local 3D editing. A central consequence of this design is that 3DEditVerse is not merely a prompt collection or an image-editing dataset lifted into 3D independently per sample. The source and edited targets are constructed to preserve correspondence, locality, and semantic harmony. This distinguishes it from datasets criticized in the same paper for lacking consistency or producing semantically awkward composites [2510.02994]. ## 2. Construction pipelines 3DEditVerse is built through two complementary pipelines: a pose-driven geometric pipeline and a foundation model-guided appearance pipeline [2510.02994]. Their complementarity is important because the first supplies naturally consistent structural variation at scale, while the second supplies semantically diverse instruction-conditioned edits. ### 2.1 Pose-driven geometric edits The geometric branch is based on publicly available **3D characters** and **animation sequences** from **Mixamo** [2510.02994]. Candidate animation frames are sampled, rendered from a canonical view, embedded with **DINOv2**, and pruned by cosine similarity to remove near-duplicate poses. This yields **4,998 unique candidate poses**. A set of **108 distinct characters** is then paired with **500 poses randomly selected** from that candidate pool, producing: $$ 108 \times 500 = 54{,}000 $$ unique 3D assets. Before–after edit pairs are formed by associating different poses of the same character. The resulting edits are geometric in the sense that they alter pose, articulation, or part configuration while maintaining identity and full 3D consistency. The benchmark text does not formalize locality mathematically for this branch, but the intended locality follows from identity-preserving articulation rather than arbitrary object replacement [2510.02994]. ### 2.2 Foundation model-guided appearance edits The second branch is a text-to-image-to-3D lifting pipeline via a cascade of foundation models [2510.02994]. It begins with the **4,585-word vocabulary** from **Tag2Text**. For each vocabulary word, **DeepSeek-R1** generates diverse descriptive prompts; the appendix example states that for “plane,” DeepSeek-R1 produces **20 descriptive prompts**, each around 20 words. These prompts are passed to **Flux.1-Dev** to produce a centered source image on a pure white background. Given the source image $I^\text{src}$, **Qwen-VL** generates an edit instruction $p^\text{edit}$, and **Flux.1-Kontext** executes the edit to obtain the target image $I^\text{tgt}$ [2510.02994]. The appendix indicates add/remove/replace-part prompt templates with emphasis on substantial visible shape transformations, although the main text characterizes this branch overall as appearance-oriented. The pipeline is further augmented with samples from **ShapeLLM-Omni** rendered from **Objaverse-XL**, whose source image and prompt pairs are also edited with Flux.1-Kontext. The resulting image pairs are then lifted into 3D with a consistency-preserving procedure rather than independently reconstructed source and target assets. This design reflects the paper’s claim that independent lifting causes geometric distortion and identity mismatch [2510.02994]. ## 3. Locality and multi-view consistency mechanisms The most technically distinctive part of 3DEditVerse is the appearance-edit lifting pipeline, which uses automatic region localization, multi-view mask propagation, 3D back-projection, and latent localized editing to ensure that source and target differ only where intended [2510.02994]. First, source and target images are separately lifted with **Trellis** to produce initial 3D assets $\hat{S}_{3D}^\text{src}$ and $\hat{S}_{3D}^\text{tgt}$. Then **Qwen-VL** is used as an open-set detector to output a 2D bounding box $B_{2D}$ for the edited region on a rendered source view. Across multiple rendered views, **SAM2** segments and tracks the edited region, producing 2D masks $\{M_{2D}^i\}$. These per-view masks are back-projected into 3D using the pinhole model. The paper gives the projection as: $$ \tilde{p}_i = K_i [R_i \mid t_i] v, \quad p_i = \left( \tfrac{\tilde{p}_{i,x}}{\tilde{p}_{i,z}}, \tfrac{\tilde{p}_{i,y}}{\tilde{p}_{i,z}} \right), $$ where $v=(x,y,z,1)^\top$ is a voxel, $K_i$ is the intrinsic matrix, and $[R_i \mid t_i]$ is the extrinsic matrix of view $i$ [2510.02994]. Multi-view support is accumulated over **70** views: $$ c(v) = \sum\nolimits_{i=1}^{N} \mathbbm{1}[p_i \in M_{2D}^i], $$ and the final 3D mask is defined by thresholding support: $$ M_{3D} = \{ v \mid c(v) \geq \tau \}. $$ This mask serves as the locality operator for latent editing. Using **Repaint** in Trellis, the paper injects noise into the source asset latent at timestep $t$, $$ z^{\text{src}}_t = \mathcal{N}\big(\hat{S}_{3D}^{\text{src}}, \sigma_t\big), $$ and then fuses source and target latents via the 3D edit mask: $$ \hat{z}_t = M_{3D} \odot z^{\text{tgt}}_t + (1 - M_{3D}) \odot z^{\text{src}}_t. $$ Inside the mask, the target latent is followed; outside the mask, the source latent is preserved [2510.02994]. This is the dataset’s main operational mechanism for encoding edit locality. Post-edit quality control then renders both the final edited asset and the initial target prediction from multiple views and filters samples with low mean **DINOv2** cosine similarity, removing inconsistent or low-quality edits. This pipeline suggests that 3DEditVerse is best understood not just as a static benchmark, but as a procedurally constructed paired-edit corpus whose supervision signal is already shaped by explicit geometric and consistency constraints. ## 4. Data content and benchmark structure The released benchmark is described as containing paired original and edited 3D assets, and its construction additionally involves rendered images, edit prompts, 2D bounding boxes, tracked 2D masks, back-projected 3D masks, and camera parameters [2510.02994]. The paper does not enumerate which intermediate fields are guaranteed to be part of the public release, but they are part of the generation and filtering process. The benchmark is explicitly compared against three earlier resources [2510.02994]: | Dataset | Train size | Test size | |---|---:|---:| | 3D-Alpaca-Editing | 52,532 | — | | CMD | 40,000 | 50 | | Edit3D-Bench | — | 300 | | 3DEditVerse | 116,309 | 1,500 | The same comparison argues that 3DEditVerse uniquely combines train/test availability with good edit region quality, consistency, and harmony [2510.02994]. Here, “consistency” refers to before–after correspondences derived from the same source identity rather than independent generation, while “harmony” refers to semantically plausible edited targets rather than concatenations of unrelated assets. The test set is manually curated through human assessment, and the paper treats that curation as a benchmark-quality control stage [2510.02994]. A plausible implication is that benchmark reliability depends not only on generation-scale automation but also on aggressive filtering and human validation. ## 5. Relationship to 3DEditFormer 3DEditVerse is introduced jointly with **3DEditFormer**, a mask-free 3D editor built by extending **Trellis** [2510.02994]. The benchmark’s role is not incidental: it supplies the paired supervision that allows the model to learn structural preservation and editable-region localization without requiring auxiliary 3D masks at inference. 3DEditFormer uses a **Dual-Guidance Attention Block**, **Multi-Stage Feature Extraction**, and **Time-Adaptive Gating**. In the $i$-th transformer block, standard self-attention and two source-guided cross-attention branches are fused as $$ h_1 = \text{SelfAttn}(\text{Norm}(x)), $$ $$ h_2 = \text{CrossAttn}_1(\text{Norm}(x), f^{(1,i)}_{3D}), \quad h_3 = \text{CrossAttn}_2(\text{Norm}(x), f^{(2,i)}_{3D}), $$ $$ h = h_1 + g_1 \odot h_2 + g_2 \odot h_3. $$ The two feature families are extracted from the frozen Trellis model at different timesteps: $$ \{f^{(1,i)}_{3D}\}_{i=1}^N = \mathcal{F}(S^{\text{src}}_{3D}, t_1, I^{\text{zero}}), \quad \{f^{(2,i)}_{3D}\}_{i=1}^N = \mathcal{F}(S^{\text{src}}_{3D}, t_2, I^{\text{tgt}}), $$ with $t_1 \approx 0$ and $t_2 \approx 1$ [2510.02994]. The timestep-dependent gates are produced by $$ (g_1, g_2) = \text{MLP}_{\text{edit}}(t_\text{embedding}). $$ The generation objective follows Trellis’s **Conditional Flow Matching** formulation: $$ \mathcal{L}(\theta_k)=\mathbb{E}_{t,\boldsymbol{x}_0,\boldsymbol{\epsilon}} \left\|\mathcal{T}^{(k)}_{\theta_k}(\boldsymbol{x}, t)-(\boldsymbol{\epsilon}-\boldsymbol{x}_0)\right\|^2_2, $$ with $$ \boldsymbol{x}(t)=(1-t)\boldsymbol{x}_0+t\boldsymbol{\epsilon}. $$ This architecture matters for understanding 3DEditVerse because the benchmark is the enabling supervision that teaches the model to separate preserved structure from edit-induced change. The claim is not merely that 3DEditVerse is large, but that its paired locality-preserving design makes mask-free structure-preserving editing learnable [2510.02994]. ## 6. Evaluation, downstream use, and later relevance On the 3DEditVerse test set, 3DEditFormer is evaluated using 3D metrics computed from **100,000 points** sampled from predicted and ground-truth meshes—**Chamfer Distance (CD)**, **Normal Consistency (NC)**, and **F1$^{0.01}$**—and 2D metrics computed from **10 fixed camera viewpoints**—**PSNR**, **SSIM**, **LPIPS**, and **DINO-I** [2510.02994]. Relative to EditP23, the paper reports large improvements on the full test set, including **CD 46.19 → 13.84**, **NC 0.689 → 0.830**, **F1$^{0.01}$ 32.33 → 64.30**, **PSNR 18.32 → 24.40**, **SSIM 0.870 → 0.918**, **LPIPS 0.158 → 0.068**, and **DINO-I 0.785 → 0.963** [2510.02994]. On the subset excluding character-animation samples, 3DEditFormer also surpasses VoxHammer on most reported metrics despite not using auxiliary 3D masks. The benchmark further serves as editing supervision for **Omni123**, a native multimodal autoregressive model that unifies text, images, and 3D in a shared discrete-token framework [2604.02289]. Omni123 includes a dedicated stage titled **“Align to Instruction-Based 3D Editing”** and uses **3DEditVerse** with **116,309 training pairs** and **1,500 curated test pairs** for source-mesh-plus-instruction to target-mesh training [2604.02289]. Its editing objective is $$ \mathcal{L}_{\text{edit}} = -\sum_{k=1}^{L_m} \log\, p_\theta\!\bigl(s'_k \mid s'_{