# M:3L Publications > An authoritative machine-readable record of M:3L publications and conference work, including status, authors, methods, datasets, paper-reported results, external records, and code where available. ## Evidence standard Benchmark claims below are tied to the associated paper and a named dataset, metric, input setting, and table. "Reported state of the art" means the associated paper reported that comparison at publication time; it is not an evergreen claim. Potential applications are kept separate from demonstrated results. ## [Z3D: Zero-Shot 3D Visual Grounding from Images](https://m3l.am/publications/z3d.md) > A universal pipeline for localizing objects in 3D scenes from multi-view images using zero-shot methods, without geometric supervision or object priors. ### Publication record - **Title:** Z3D: Zero-Shot 3D Visual Grounding from Images - **Authors:** Nikita Drozdov, Andrey Lemeshko, Nikita Gavrilov, Anton Konushin, Danila Rukhovich, Maksim Kolodiazhnyi - **Year:** 2026 - **Status:** Accepted at ACL 2026 - **Venue:** ACL 2026 - **Conference ranking:** A*-ranked conference - **Research area:** 3D Intelligence - **First arXiv submission:** 2026-02-03 ### Abstract / summary Z3D is a universal zero-shot 3D visual-grounding pipeline that localizes the object described by a natural-language query. It operates on multi-view images, can optionally incorporate camera poses and depth maps, and avoids task-specific geometric supervision or predefined object priors. ### Methodology - VLM-assisted image and view selection - Prompt-based 2D segmentation with SAM3-Agent - Zero-shot 3D proposal generation with MaskClustering - 2D-to-3D lifting and multi-view proposal aggregation - Optional DUSt3R reconstruction for posed or unposed imagery ### Datasets and benchmarks - ScanRefer - Nr3D ### Principal results - The paper reports state-of-the-art performance among zero-shot methods on ScanRefer and Nr3D at publication time. - In the full ScanRefer validation setting without bounding-box or text supervision, Z3D reports overall Acc@0.25 of 54.2 and Acc@0.5 of 46.0. - On Nr3D, Z3D reports overall top-1 accuracy of 54.8 in the paper's main zero-shot comparison. - The same pipeline supports point-cloud, posed-RGB, and fully unposed-RGB inputs. ### Benchmark evidence | Benchmark | Task | Input | Metric | Result | Evidence | | --- | --- | --- | --- | --- | --- | | ScanRefer | Zero-shot 3D visual grounding | Ground-truth point cloud; no bounding-box or text supervision | Overall Acc@0.25 / Acc@0.5 | 54.2 / 46.0 | Z3D paper, Table 1; Reported as state of the art among zero-shot approaches in the associated paper. | | ScanRefer | Zero-shot 3D visual grounding | Posed multi-view RGB | Overall Acc@0.25 / Acc@0.5 | 42.8 / 24.8 | Z3D paper, Table 1 | | ScanRefer | Zero-shot 3D visual grounding | Unposed multi-view RGB | Overall Acc@0.25 / Acc@0.5 | 31.2 / 12.9 | Z3D paper, Table 1 | | Nr3D | Zero-shot 3D visual grounding | Depth-aware setting | Overall top-1 accuracy | 54.8 | Z3D paper, Table 2; The strongest zero-shot result in the paper's main Nr3D table was 54.8 versus 54.3 for SPAZER. | Metrics are reported exactly for the stated input and evaluation setting; rows from different settings should not be treated as directly interchangeable. ### Technologies and methods - PyTorch - CLIP - SAM3 - Qwen vision-language models - DUSt3R ### Related M:3L research - [3D Intelligence](https://m3l.am/research/3d-intelligence.md) - [CAD Intelligence](https://m3l.am/research/cad-intelligence.md) ### External resources - [arXiv record](https://arxiv.org/abs/2602.03361) - [DOI](https://doi.org/10.48550/arXiv.2602.03361) - [Code repository](https://github.com/col14m/z3d) ### Evidence boundary Benchmark claims and numerical values above are attributed to the associated paper and its stated protocol. Potential applications are not evidence of completed deployments. ## [Zoo3D: Zero-Shot 3D Object Detection at Scene Level](https://m3l.am/publications/zoo3d.md) > The first training-free 3D object detection framework. It constructs 3D bounding boxes through graph clustering of 2D instance masks and assigns open-vocabulary semantic labels. ### Publication record - **Title:** Zoo3D: Zero-Shot 3D Object Detection at Scene Level - **Authors:** Andrey Lemeshko, Bulat Gabdullin, Nikita Drozdov, Anton Konushin, Danila Rukhovich, Maksim Kolodiazhnyi - **Year:** 2026 - **Status:** Accepted at CVPR 2026 - **Venue:** CVPR 2026 - **Conference ranking:** A*-ranked conference - **Research area:** 3D Intelligence - **First arXiv submission:** 2025-11-25 ### Abstract / summary Zoo3D introduces scene-level zero-shot 3D object detection. Its training-free Zoo3D0 mode constructs class-agnostic 3D boxes from 2D instance masks and assigns open-vocabulary labels; Zoo3D1 then trains a class-agnostic detector on Zoo3D0-generated pseudo-labels. ### Methodology - 2D instance-mask extraction across scene views - Graph clustering for class-agnostic 3D bounding-box construction - Best-view selection and view-consensus mask generation - Open-vocabulary visual-language semantic labeling - Zero-shot Zoo3D0 and pseudo-label-based self-supervised Zoo3D1 modes - DUSt3R-based support for posed and unposed image inputs ### Datasets and benchmarks - ScanNet20 - ScanNet60 - ScanNet200 - ARKitScenes ### Principal results - The paper reports state-of-the-art open-vocabulary 3D-detection results across ScanNet-based benchmarks at publication time. - On ScanNet200 with point-cloud inputs, Zoo3D0 reports mAP@0.25/mAP@0.5 of 21.1/14.1 without training, while Zoo3D1 reports 23.5/15.2. - On ARKitScenes, the paper evaluates point-cloud and unposed-image tracks and reports mAP at IoU thresholds of 0.25 and 0.5. ### Benchmark evidence | Benchmark | Task | Input | Metric | Result | Evidence | | --- | --- | --- | --- | --- | --- | | ScanNet200 | Open-vocabulary 3D object detection (Zoo3D0) | Point cloud and posed scene images; training-free | mAP@0.25 / mAP@0.5 | 21.1 / 14.1 | Zoo3D paper, Table 2 | | ScanNet200 | Open-vocabulary 3D object detection (Zoo3D1) | Point cloud and posed scene images; self-supervised pseudo-label training | mAP@0.25 / mAP@0.5 | 23.5 / 15.2 | Zoo3D paper, Table 2 | | ScanNet200 | Open-vocabulary 3D object detection (Zoo3D0 / Zoo3D1) | Fully unposed multi-view images | mAP@0.25 / mAP@0.5 | 8.3 / 2.9 (Zoo3D0); 10.7 / 3.8 (Zoo3D1) | Zoo3D paper, Table 2 | | ARKitScenes | Open-vocabulary 3D object detection (Zoo3D0 / Zoo3D1) | Point cloud and posed images | mAP@0.25 / mAP@0.5 | 24.4 / 11.0 (Zoo3D0); 34.2 / 24.2 (Zoo3D1) | Zoo3D paper, Table 15 | | ARKitScenes | Open-vocabulary 3D object detection (Zoo3D0 / Zoo3D1) | Unposed multi-view images | mAP@0.25 / mAP@0.5 | 13.0 / 2.6 (Zoo3D0); 16.1 / 3.5 (Zoo3D1) | Zoo3D paper, Table 15 | Metrics are reported exactly for the stated input and evaluation setting; rows from different settings should not be treated as directly interchangeable. ### Technologies and methods - CLIP - SAM 2.1 - MaskClustering - DUSt3R - Sparse 3D detection ### Related M:3L research - [3D Intelligence](https://m3l.am/research/3d-intelligence.md) - [CAD Intelligence](https://m3l.am/research/cad-intelligence.md) ### External resources - [arXiv record](https://arxiv.org/abs/2511.20253) - [DOI](https://doi.org/10.48550/arXiv.2511.20253) - [Code repository](https://github.com/col14m/zoo3d) ### Evidence boundary Benchmark claims and numerical values above are attributed to the associated paper and its stated protocol. Potential applications are not evidence of completed deployments. ## [TUN3D: Towards Real-World Scene Understanding from Unposed Images](https://m3l.am/publications/tun3d.md) > The first method for joint layout estimation and 3D detection without camera poses or depth supervision, designed to work with point clouds, posed images, or unposed RGB captures. ### Publication record - **Title:** TUN3D: Towards Real-World Scene Understanding from Unposed Images - **Authors:** Anton Konushin, Nikita Drozdov, Bulat Gabdullin, Alexey Zakharov, Anna Vorontsova, Danila Rukhovich, Maksim Kolodiazhnyi - **Year:** 2026 - **Status:** Accepted at ICRA 2026 - **Venue:** ICRA 2026 - **Conference ranking:** A*-ranked conference - **Research area:** 3D Intelligence - **First arXiv submission:** 2025-09-23 ### Abstract / summary TUN3D jointly estimates indoor scene layouts and detects 3D objects from point clouds, posed images, or fully unposed multi-view RGB. In the unposed setting it does not require known camera poses, camera calibration, or externally provided depth as input. ### Methodology - Lightweight sparse-convolutional scene backbone - Dedicated 3D object-detection and wall-layout heads - A bird's-eye-view parametric wall representation with two 2D corner offsets and wall height - DUSt3R reconstruction for posed or fully unposed image collections - Joint optimization of object classification, 3D boxes, wall classification, and wall parameters ### Datasets and benchmarks - ScanNet - S3DIS - ARKitScenes - Structured3D ### Principal results - The paper presents the first reported joint layout-estimation and 3D-detection results on real scans from fully unposed images. - On unposed ScanNet imagery, TUN3D reports layout F1 of 46.5, object-detection mAP@0.25 of 44.0, and mAP@0.5 of 20.7. - With ground-truth point clouds, the paper reports state-of-the-art layout results across ScanNet, S3DIS, ARKitScenes, and Structured3D at publication time. ### Benchmark evidence | Benchmark | Task | Input | Metric | Result | Evidence | | --- | --- | --- | --- | --- | --- | | ScanNet | Joint layout estimation and 3D object detection | Unposed multi-view RGB; no known poses or external depth | Layout F1 / detection mAP@0.25 / mAP@0.5 | 46.5 / 44.0 / 20.7 | TUN3D paper, Table I | | S3DIS | Joint layout estimation and 3D object detection | Unposed multi-view RGB; no known poses or external depth | Layout F1 / detection mAP@0.25 / mAP@0.5 | 20.8 / 11.0 / 2.2 | TUN3D paper, Table I | | ScanNet | Joint layout estimation and 3D object detection | Posed multi-view RGB; no ground-truth depth | Layout F1 / detection mAP@0.25 / mAP@0.5 | 55.2 / 57.4 / 35.6 | TUN3D paper, Table I | | ARKitScenes | Layout estimation | Ground-truth point cloud; cross-dataset evaluation | Layout F1 | 30.3 | TUN3D paper, Table II; The paper reports 25.9 for Omni-PQ under the same table protocol. | | Structured3D | Layout estimation and 3D object detection | Ground-truth point cloud | Layout F1@0.25/F1@0.5; detection F1@0.25/F1@0.5 | 90.5/89.6; 73.9/65.4 | TUN3D paper, Table III | Metrics are reported exactly for the stated input and evaluation setting; rows from different settings should not be treated as directly interchangeable. ### Technologies and methods - Sparse convolutional networks - DUSt3R - TSDF fusion - Finite-dimensional parametric wall models ### Related M:3L research - [3D Intelligence](https://m3l.am/research/3d-intelligence.md) - [CAD Intelligence](https://m3l.am/research/cad-intelligence.md) - [Agentic AI](https://m3l.am/research/agentic-ai.md) ### External resources - [arXiv record](https://arxiv.org/abs/2509.21388) - [DOI](https://doi.org/10.48550/arXiv.2509.21388) - [Code repository](https://github.com/col14m/tun3d) - [Project page](https://bulatko.github.io/tun3d/) ### Evidence boundary Benchmark claims and numerical values above are attributed to the associated paper and its stated protocol. Potential applications are not evidence of completed deployments. ## [cadrille: Multi-modal CAD Reconstruction with Online Reinforcement Learning](https://m3l.am/publications/cadrille.md) > A vision-language model that processes point clouds, multi-view images, and text to generate editable CAD programs. It is the first work to apply online reinforcement learning to CAD reconstruction and achieves a near-zero invalidity ratio. ### Publication record - **Title:** cadrille: Multi-modal CAD Reconstruction with Online Reinforcement Learning - **Authors:** Maksim Kolodiazhnyi, Denis Tarasov, Dmitrii Zhemchuzhnikov, Alexander Nikulin, Ilya Zisman, Anna Vorontsova, Anton Konushin, Vladislav Kurenkov, Danila Rukhovich - **Year:** 2026 - **Status:** Accepted as an Oral at ICLR 2026 - **Venue:** ICLR 2026 Oral - **Conference ranking:** A*-ranked conference - **Research area:** CAD Intelligence - **First arXiv submission:** 2025-05-28 ### Abstract / summary cadrille is a multimodal CAD-reconstruction model that processes point clouds, multi-view images, or text and generates executable Python-based CadQuery programs. It combines supervised fine-tuning with online reinforcement learning driven by programmatic CAD feedback. ### Methodology - One Qwen2-VL-based vision-language architecture for point clouds, images, and text - Supervised fine-tuning on DeepCAD, CAD-Recode, and text-enriched CAD data - Executable CadQuery program generation - Online Dr. CPPO reinforcement learning with programmatically computed geometric feedback - Evaluation by Chamfer Distance, volumetric Intersection over Union, and Invalidity Ratio ### Datasets and benchmarks - DeepCAD - Fusion360 - CC3D - Omni-CAD - CAD-Recode - Text2CAD ### Principal results - The paper reports state-of-the-art multimodal CAD-reconstruction results on DeepCAD, Fusion360, and real-world CC3D at publication time. - Online RL reduces invalid CAD programs to 0.0-0.2% across the main point-cloud and multi-view-image benchmark rows. - On multi-view images, the RL model reports IoU of 92.2% on DeepCAD, 84.6% on Fusion360, and 65.0% on CC3D. - On point clouds, the RL model reports IoU of 90.2% on DeepCAD, 85.0% on Fusion360, and 67.9% on CC3D. ### Benchmark evidence | Benchmark | Task | Input | Metric | Result | Evidence | | --- | --- | --- | --- | --- | --- | | DeepCAD | CAD reconstruction | Multi-view images; online Dr. CPPO RL | Median Chamfer Distance x10^3 / IoU % / Invalidity Ratio % | 0.17 / 92.2 / 0.0 | cadrille paper, Table 2 | | Fusion360 | CAD reconstruction | Multi-view images; online Dr. CPPO RL | Median Chamfer Distance x10^3 / IoU % / Invalidity Ratio % | 0.17 / 84.6 / 0.0 | cadrille paper, Table 2 | | CC3D | CAD reconstruction from real-world scanned models | Multi-view images; online Dr. CPPO RL | Median Chamfer Distance x10^3 / IoU % / Invalidity Ratio % | 0.57 / 65.0 / 0.1 | cadrille paper, Table 2 | | DeepCAD | CAD reconstruction | Point cloud; online Dr. CPPO RL | Median Chamfer Distance x10^3 / IoU % / Invalidity Ratio % | 0.17 / 90.2 / 0.0 | cadrille paper, Table 3 | | Fusion360 | CAD reconstruction | Point cloud; online Dr. CPPO RL | Median Chamfer Distance x10^3 / IoU % / Invalidity Ratio % | 0.17 / 85.0 / 0.2 | cadrille paper, Table 3 | | CC3D | CAD reconstruction from real-world scanned models | Point cloud; online Dr. CPPO RL | Median Chamfer Distance x10^3 / IoU % / Invalidity Ratio % | 0.47 / 67.9 / 0.2 | cadrille paper, Table 3 | | DeepCAD | Text-to-CAD reconstruction | Text; supervised fine-tuning | Median Chamfer Distance x10^3 / IoU % / Invalidity Ratio % | 0.20 / 82.1 / 1.4 | cadrille paper, Table 1 | Metrics are reported exactly for the stated input and evaluation setting; rows from different settings should not be treated as directly interchangeable. ### Technologies and methods - PyTorch - PyTorch3D - Transformers - Qwen2-VL - CadQuery - Online reinforcement learning ### Related M:3L research - [CAD Intelligence](https://m3l.am/research/cad-intelligence.md) - [3D Intelligence](https://m3l.am/research/3d-intelligence.md) - [Agentic AI](https://m3l.am/research/agentic-ai.md) ### External resources - [arXiv record](https://arxiv.org/abs/2505.22914) - [DOI](https://doi.org/10.48550/arXiv.2505.22914) - [Code repository](https://github.com/col14m/cadrille) ### Evidence boundary Benchmark claims and numerical values above are attributed to the associated paper and its stated protocol. Potential applications are not evidence of completed deployments. ## [Drivers of Heritage Site Change: A Remote Evaluation Framework](https://m3l.am/publications/heritage-site-change.md) > A dynamic, multi-level assessment using computer-vision analysis of historic satellite imagery over monumental heritage sites of Armenia. ### Publication record - **Title:** Drivers of Heritage Site Change: A Remote Evaluation Framework - **Authors:** Not listed in the current public record - **Year:** 2025 - **Status:** Presented at ICAHCA 2025 - **Venue:** ICAHCA 2025, Vienna, Austria - **Research area:** Computer Vision for Heritage Monitoring ### Abstract / summary A dynamic, multi-level framework for evaluating change drivers at Armenian monumental heritage sites using computer-vision analysis of historic satellite imagery. ### Methodology - Computer vision - Historic satellite imagery - Multi-level change assessment ### Datasets and benchmarks - Historic satellite imagery of monumental heritage sites in Armenia ### Principal results - Presented as research at ICAHCA 2025 in Vienna, Austria. ### Technologies and methods - Computer vision - Remote sensing ### Evidence boundary Benchmark claims and numerical values above are attributed to the associated paper and its stated protocol. Potential applications are not evidence of completed deployments.