Pith. sign in

REVIEW 4 major objections 4 minor 48 references

CausalSplat: Towards Comprehensive Hierarchical Reasoning in 3D Gaussian Splatting

T0 review · 4 major / 4 minor · reviewed 2026-08-12 · deepseek-v4-flash

Pith's one-line read Routing implicit 3D instructions through a scene graph and a vision-language model yields state-of-the-art reasoning segmentation in Gaussian-splat scenes.

desk verdict Genuinely new task and taxonomy, plausible pipeline with one solid external result, but the headline SOTA rests on unreleased benchmarks that need to be opened before the numbers can be trusted. read the letter →

arxiv 2608.11150 v1 pith:YGOB2C24 submitted 2026-08-11 cs.CV

classification cs.CV
keywords reasoning3DsegmentationGaussianSplattingscenegraphvision-languagemodelcounterfactualreferringopen-vocabularybenchmark
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper introduces a new task, 'reasoning 3D Gaussian segmentation,' in which a model must turn an implicit natural-language instruction into a precise 3D object mask inside a Gaussian-splatted scene. To measure it, the authors build two benchmarks, Causal-LERF (2D evaluation) and Causal-ScanNet (3D evaluation), with 231 instructions spanning spatial, commonsense, affordance, and counterfactual reasoning. They report that existing open-vocabulary, referring, and reasoning baselines perform poorly, with the strongest prior method reaching only 23.6 mIoU on Causal-LERF and 5.1 mIoU on Causal-ScanNet. The proposed CausalSplat framework, which builds a multimodal semantic scene graph from the Gaussian point cloud and then lets a vision-language model reason over that graph, reaches 47.0 and 14.9 mIoU respectively. If the benchmarks are fair, the paper's central claim is that explicit structural perception (scene graph) and implicit logical inference (VLM) should be decoupled, and that this separation is what enables the large margin.

What carries the argument

The central object is the multimodal semantic scene graph, where each node is a 3D entity described by its centroid, bounding-box span, associated 2D masks, and a VLM-generated attribute tuple (category, visual features, affordances), and each edge encodes a spatial relation such as 'supported by' or 'near' using a scale-adaptive threshold that depends on object size and scene span. This graph is the substrate for a three-stage reasoning pipeline: instruction parsing (deciding whether topology is needed), topological search along directed edges from an anchor node, and decision output that fuses semantic and spatial constraints. The machinery also includes a semantic field construction step that suppresses 2D mask boundary noise by spatially weighted feature aggregation and contrastive optimization before clustering masks into entities, which the ablation study shows is necessary to reach the reported accuracy.

What would settle it

Take the released (or independently regenerated) Causal-LERF and Causal-ScanNet instructions and run the CausalSplat pipeline with the same scene-graph construction but a different VLM (for example, an open-weights model from a different family) as the graph builder and query parser; if the margin over LUDVIG collapses below a few mIoU points, the claimed structural advantage is VLM-specific rather than a property of the scene-graph design. A second check: verify that instruction templates prepared before seeing the method do not contain the same wordings that the method's VLM was fine-tuned on, which would indicate benchmark leakage.

Watch

Extended reading notes

Core claim

The paper claims that the bottleneck in 3D reasoning segmentation is not visual feature quality but the conflation of two distinct jobs: perceiving what objects exist and how they sit in space, versus inferring what an ambiguous instruction means. CausalSplat separates these by first lifting 2D masks into 3D entity clusters with a contrastively optimized semantic field, then organizing those entities into a scene graph whose nodes carry geometric, visual, and functional attributes and whose edges encode scale-adaptive spatial relations. A vision-language model (specifically Qwen3-VL-30B-A3B-Instruct) parses the instruction, searches the graph topologically, and outputs a target entity. On the two new benchmarks the method outperforms all baselines on every reasoning level, with the biggest gains on spatial and counterfactual reasoning; on standard referring and open-vocabulary benchmarks it also sets new bests among point-based methods, at 36.1 and 51.3 mIoU.

Load-bearing premise

The new benchmarks must be valid measures of reasoning rather than artifacts of how the instructions were generated; the paper does not identify the generation model, show the templates, report annotator agreement, or release the data.

Editorial extensions

If this is right

  • R3DGS-style feature-alignment methods are not sufficient for implicit instructions; a structured scene representation plus a VLM is the current leading recipe, on the authors' benchmarks.
  • The four-level taxonomy (spatial, commonsense, affordance, counterfactual) provides a reusable evaluation grid for 3D reasoning, and the per-level results show the method beats every baseline on every level, not just on aggregate.
  • Because the same pipeline transfers without redesign to referring segmentation (Ref-LERF) and open-vocabulary point-based segmentation (LERF), the decoupling appears to be a general strategy rather than a benchmark-specific trick.
  • Ablations indicate that spatial weighting in the semantic field, multimodal nodes plus topological edges in the graph, and the complete chain-of-thought prompt each contribute; dropping any one of them costs several mIoU points.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Editorial inference: the reported margin likely depends on the choice of VLM; substituting a different vision-language model while keeping the scene graph fixed would separate framework gains from VLM capability, an experiment the paper does not run.
  • Editorial inference: because the instruction-generation LLM is not identified and the templates and annotator agreement are not reported, the benchmarks may be partially aligned with the authors' own pipeline; a held-out instruction set generated by a different model with reported inter-annotator agreement would settle this.
  • Editorial inference: the same scene-graph-plus-VLM decoupling could be ported to other 3D representations such as NeRF or textured meshes, where explicit structure is less natural to extract.
  • Editorial inference: the failure analysis suggests that the framework's ceiling is set by SAM mask quality and VLM parsing of nested spatial constraints, so improving those two components should yield the next jump, not a change of scene-graph design.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 4 minor

Summary. The paper introduces a new task, "Reasoning 3D Gaussian Segmentation," where a model must segment a 3D Gaussian scene from implicit, multi-hop natural language queries. It defines a four-level reasoning taxonomy (spatial, commonsense, affordance, predictive/counterfactual), constructs two new benchmarks (Causal-LERF and Causal-ScanNet) with 231 instructions over 14 scenes, and proposes CausalSplat, a pipeline that builds a semantic feature field, constructs a multimodal 3D scene graph, and uses a VLM (Qwen3-VL-30B-A3B-Instruct) to parse instructions and search the graph. The central empirical claim is state-of-the-art performance on the two new benchmarks (47.0 vs 23.6 mIoU on Causal-LERF; 14.9 vs 5.1 on Causal-ScanNet) plus strong generalization on the external Ref-LERF (36.1 vs 29.2) and LERF open-vocabulary benchmarks.

Significance. If the benchmark validity and the reported numbers hold, the paper would be a useful step toward reasoning-level 3D scene understanding: it names a concrete task, provides a hierarchical evaluation instrument, and demonstrates that a scene-graph-plus-VLM decomposition can outperform direct feature-alignment baselines on implicit queries. The external Ref-LERF result gives partial independent support for the method's localization ability. The paper also reports a full ablation of the main components and a runtime/memory profile, which are useful for reproducibility. However, the core significance rests on two self-constructed benchmarks whose generation process, data, and evaluation protocol are not disclosed; until those are made available and the evaluation is shown to be robust, the claimed state of the art on the new task cannot be independently assessed.

major comments (4)
  1. [§3.3 and Tables 2–3] The validity of the new benchmarks is load-bearing but not sufficiently documented. §3.3 states that instructions are generated by "an LLM" from ground-truth annotations and preset templates and then reviewed by four annotators, but it does not identify the generation model, show the templates, report annotator agreement, or release the instructions and annotations. Since the headline SOTA claims (47.0 vs 23.6 on Causal-LERF and 14.9 vs 5.1 on Causal-ScanNet) are measured only on these benchmarks, the lack of data release and of a detailed generation protocol makes it impossible to check for annotation artifacts, ambiguous queries, or distributional overlap with the method's VLM. The authors should release the benchmark data, templates, generation prompts, and per-instruction annotations, and report inter-annotator agreement.
  2. [§4.3, §5.1, and §3.3] There is a potential circularity between benchmark construction and method design: the method's reasoning module uses Qwen3-VL to parse query semantics and to build scene-graph attributes, while §3.3 uses an unnamed LLM to generate the benchmark instructions. If the instruction generator is the same model family, or if the preset templates encode Qwen3-VL's phrasing preferences, the large margins on the new benchmarks may reflect alignment between the query distribution and the method's internal VLM rather than general reasoning ability. The authors should disclose the generation model, compare against a benchmark generated with a different model or human-written queries, and report results separately by generation source.
  3. [§5.1, Table 9, and §A.4] Hyperparameters appear to be selected on the same benchmark that reports the headline results. The vertical-edge threshold τ in Eq. (9) is ablated on Causal-LERF in Table 9, and the value τ=0.3 is chosen because it maximizes the Causal-LERF mIoU. With only 158 queries in Causal-LERF and no separate development/test split, such selection can inflate the reported margin. The authors should either fix hyperparameters before evaluation, use cross-validation, or report sensitivity of the main results to τ and the other free parameters (ω_min, contrastive margin m, hard-negative threshold τ in Eq. (7), and the coefficients in Eq. (9)).
  4. [§5.2 and §A.3] The quantitative evaluation lacks error bars or statistical significance measures, and the relationship between the 231-instruction benchmarks and the expanded 2355-instruction evaluation in Table 8 is unclear. Table 8 reports different average scores (e.g., Causal-LERF Avg 40.5 vs the 47.0 in Table 2) without specifying whether the instruction set, scene splits, or annotation process differ from the main benchmarks. The paper should report variance across runs or evaluation subsets, and clarify whether the expanded set is a separate held-out test set or an extension of the same benchmarks on which hyperparameters were tuned.
minor comments (4)
  1. [§4.1 and §4.2] The symbol τ is used for two different hyperparameters: the hard-negative sampling threshold in Eq. (7) and the vertical-displacement threshold in Eq. (9), and §5.1 then lists both τ=0.5 and 0.3 without distinguishing them. Please use distinct symbols, e.g., τ_neg and τ_vert.
  2. [§4.1 and References] The citation numbers in Section 4.1 appear inconsistent with the reference list: the 3DGS rasterization pipeline is cited as [10] but the 3DGS reference is [15], and SAM is cited as [13] but the SAM reference is [18]. Please recheck all citation numbers.
  3. [Table 6 and §5.5] The ablation row "Random Weighting" is not defined in the method section; please specify how the random weighting is generated (e.g., random spatial weights per mask) so that the ablation is reproducible.
  4. [§A.3 and Table 8] The sentence "These results validate our claims of zero-shot capabilities" is ambiguous because the method does not train on the benchmark instructions; please clarify what is meant by zero-shot here and how the expanded dataset was constructed.

Circularity Check

0 steps flagged · score 0.0 of 10

No demonstrated circularity: the reasoning-benchmark results are under-disclosed and hyperparameter selection is done on the test split, but the paper's central generalization claims are supported by independent external benchmarks, and no load-bearing equation or self-citation reduces the result to its inputs.

full rationale

The paper contains no formal derivation chain whose conclusion is presupposed by its premises; CausalSplat is an engineered pipeline of SAM masks, contrastive feature optimization, HDBSCAN clustering, a scene graph, and Qwen3-VL reasoning. The main reasoning results are measured on the authors' own Causal-LERF and Causal-ScanNet benchmarks, which are generated by an unidentified LLM from ground-truth annotations and templates with four-annotator review (Section 3.3). This is a benchmark-construction and disclosure risk, not a demonstrated circularity: nothing in the text identifies the instruction generator as Qwen3-VL or otherwise forces the benchmark query distribution to coincide with CausalSplat's internal reasoning. Similarly, the vertical-edge threshold tau is selected on Causal-LERF (Table 9) and the same split is reported as the headline 47.0 mIoU; that is a test-set-overfitting concern, but the reported number is not a prediction obtained by construction from the fitted parameter. The external Ref-LERF result (36.1 vs ReferSplat's 29.2, Table 4) and LERF open-vocabulary result (51.3 point-based SOTA, Table 5) are independent evidence for the method's transferable segmentation ability. Self-citations (References 6, 7, 8) appear only in related-work context and are not load-bearing. Appendix B explicitly discloses failure modes rather than hiding them. For these reasons, no specific circular step can be exhibited under the required standard, and the appropriate finding is no significant circularity; the benchmark validity gap should be treated as a correctness/evidence concern rather than a circularity score.

Assumptions & free parameters 5 free parameters · 5 assumptions · 0 invented entities

The method's performance depends on several hand-set constants and on the reliability of external components (SAM, Qwen3-VL, HDBSCAN). Most importantly, the evaluation depends on the assumption that LLM-generated, annotator-approved queries are unbiased ground truth for four reasoning abilities, an assumption the paper does not substantiate with agreement statistics or a data release.

free parameters (5)
  • omega_min (spatial weighting floor) = 0.4
    Minimum spatial weight in Eq. 2, set by hand in Section 5.1; controls how strongly mask-border pixels contribute to mask-averaged features.
  • contrastive margin m = 0.03
    Margin in the negative repulsion loss Eq. 6, set by hand in Section 5.1; changes how separable positive and negative features become.
  • hard negative sampling threshold tau = 0.5
    Hard-negative threshold in Eq. 7, set by hand in Section 5.1; affects which negative pixels are sampled during contrastive optimization.
  • vertical edge threshold tau = 0.3
    Principal-axis displacement ratio for scene graph edges; selected on Causal-LERF in Appendix Table 9 because it gives the best mIoU (47.0), meaning it is tuned on the evaluation benchmark.
  • horizontal tolerance coefficients (0.55, 0.08, 0.35) = 0.55, 0.08, 0.35
    Coefficients in Eq. 9 for the dynamic horizontal distance limit D_xy_limit; no ablation is reported, but these constants directly determine which spatial edges exist in the scene graph.
assumptions (5)
  • domain assumption Each HDBSCAN cluster produced from mask-averaged features corresponds to one physical object entity.
    Section 4.1 assigns every Gaussian point to an entity ID via clustering. Appendix B admits that extreme SAM noise can merge invalid masks into target clusters, so this one-to-one correspondence can fail.
  • domain assumption SAM 2D segmentation masks provide reliable cross-view instance evidence.
    Semantic field construction starts from SAM masks in Section 4.1. Appendix B lists large textureless backgrounds and cluttered occlusion as failure modes where SAM produces persistent errors.
  • domain assumption Qwen3-VL's knowledge and parsing are sufficient to resolve commonsense, affordance, spatial, and counterfactual queries from the scene graph.
    Section 4.3 uses Qwen3-VL-30B-A3B-Instruct for graph attribute extraction and query parsing. Appendix B documents VLM semantic matching failures on ultra-complex spatial constraints.
  • domain assumption The LLM-generated, four-annotator-approved instructions are valid ground truth for the four claimed reasoning abilities.
    Section 3.3 describes instruction generation and approval but gives no inter-annotator agreement, no annotation release, and no protocol preventing queries that favor the authors' VLM-based pipeline.
  • domain assumption The alpha-blending feature rendering in Eq. 1 preserves semantic features across views the way color rendering preserves appearance.
    The method assumes that splatting semantic feature vectors with the same alpha-blending weights as colors yields a rendered feature map whose similarities are meaningful for matching and clustering.

how reviews work

0 comments
Cite this review

Pith. "Pith review of CausalSplat: Towards Comprehensive Hierarchical Reasoning in 3D Gaussian Splatting." pith.science (2026). https://pith.science/paper/YGOB2C24

@misc{pith2026260811150,
  author       = {Pith},
  title        = {Pith review of: CausalSplat: Towards Comprehensive Hierarchical Reasoning in 3D Gaussian Splatting},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/YGOB2C24}},
  note         = {Machine review of arXiv:2608.11150}
}
read the original abstract

While 3D Gaussian Splatting (3DGS) has advanced open vocabulary scene understanding, existing methods remain confined to explicit queries. They struggle to interpret implicit intents, complex spatial constraints, and commonsense reasoning required for practical embodied interactions. To address this gap, we introduce the task of reasoning 3D Gaussian segmentation and construct two benchmarks, Causal-LERF and Causal-ScanNet. These benchmarks systematically evaluate commonsense, spatial, affordance, and counterfactual reasoning. Evaluations reveal that current state of the art methods perform poorly on these reasoning challenges. Therefore, we propose CausalSplat, a framework that integrates vision-language models with 3D scene graphs to disentangle explicit structural perception from implicit logical inference. Extensive experiments demonstrate that CausalSplat achieves state of the art performance on our reasoning benchmarks while showing strong generalizability on standard referring and open vocabulary 3D segmentation tasks. Project Page: https://jiayuding031020.github.io/CausalSplat

Figures

Figures reproduced from arXiv: 2608.11150 by the authors.

Figure 1
Figure 1. Overview of our reasoning segmentation paradigm and benchmark. Left: Evolution of 3DGS segmentation from basic [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Overview of the proposed CausalSplat pipeline. (a) Semantic Field Construction. 2D masks are extracted via SAM, and [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Qualitative comparison on the Causal-LERF dataset. [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Qualitative comparison on the Causal-ScanNet dataset. [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: Qualitative segmentation results on the Ref-LERF dataset. [PITH_FULL_IMAGE:figures/full_fig_p012_5.png]
Figure 6
Figure 6. Figure 6: Qualitative search results on the LERF dataset. [PITH_FULL_IMAGE:figures/full_fig_p013_6.png]
Figure 7
Figure 7. Figure 7: Core failure modes of our method. (a) Extreme seg [PITH_FULL_IMAGE:figures/full_fig_p013_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

48 extracted references · 27 canonical work pages

  1. [1]

    Daichi Azuma, Taiki Miyanishi, Shuhei Kurita, and Motoaki Kawanabe. 2022. Scanqa: 3d question answering for spatial scene understanding. In2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). IEEE, 19107– 19117

  2. [2]

    Dave Zhenyu Chen, Angel X Chang, and Matthias Nießner. 2020. Scanrefer: 3d object localization in rgb-d scans using natural language. InEuropean conference on computer vision. Springer, 202–221

  3. [3]

    Tianrun Chen, Chunan Yu, Jing Li, Jianqi Zhang, Lanyun Zhu, Deyi Ji, Yong Zhang, Ying Zang, Zejian Li, and Lingyun Sun. 2024. Reasoning3D–Grounding and Reasoning in 3D: Fine-Grained Zero-Shot Open-Vocabulary 3D Reasoning Part Segmentation via Large Vision-Language Models.arXiv preprint arXiv:2405.19326 (2024)

  4. [4]

    Jiahuan Cheng, Jan-Nico Zaech, Luc Van Gool, and Danda Pani Paudel. 2025. Occam’s LGS: An Efficient Approach for Language Gaussian Splatting. In36th British Machine Vision Conference 2025, BMVC 2025, Sheffield, UK, November 24-27,

  5. [5]

    Angela Dai, Angel X Chang, Manolis Savva, Maciej Halber, Thomas Funkhouser, and Matthias Nießner. 2017. Scannet: Richly-annotated 3d reconstructions of indoor scenes. InProceedings of the IEEE conference on computer vision and pattern recognition. 5828–5839

  6. [6]

    Jiayu Ding, Xinpeng Liu, Zhiyi Pan, Shiqiang Long, and Ge Li. 2026. ExtrinSplat: Decoupling Geometry and Semantics for Open-Vocabulary Understanding in 3D Gaussian Splatting. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 31019–31028

  7. [7]

    Jiayu Ding, Meilu Song, Xiaoyi Zhang, Hongbo Jin, Yichen Jin, and Xiangtian Si

  8. [8]

    Jiayu Ding, Haoran Tang, Hongbo Jin, Wei Gao, and Ge Li. 2026. 3D Instruction Ambiguity Detection.arXiv preprint arXiv:2601.05991(2026)

Show all 48 references
  1. [9]

    Agrim Gupta, Silvio Savarese, Surya Ganguli, and Li Fei-Fei. 2021. Embodied intelligence via learning and evolution.Nature communications12, 1 (2021), 5721. Publisher: Nature Publishing Group UK London

  2. [10]

    Shuting He, Henghui Ding, Xudong Jiang, and Bihan Wen. 2024. Segpoint: Segment any point cloud via large language model. InEuropean Conference on Computer Vision. Springer, 349–367

  3. [11]

    Shuting He, Guangquan Jie, Changshuo Wang, Yun Zhou, Shuming Hu, Guanbin Li, and Henghui Ding. 2025. ReferSplat: Referring Segmentation in 3D Gaussian Splatting. InInternational Conference on Machine Learning (ICML)

  4. [12]

    Kuan-Chih Huang, Xiangtai Li, Lu Qi, Shuicheng Yan, and Ming-Hsuan Yang

  5. [13]

    Kim Jun-Seong, GeonU Kim, Kim Yu-Ji, Yu-Chiang Frank Wang, Jaesung Choe, and Tae-Hyun Oh. 2025. Dr. splat: Directly referring 3D gaussian splatting via direct language embedding registration. InProceedings of the Computer Vision and Pattern Recognition Conference. 14137–14146

  6. [14]

    Amrin Kareem, Jean Lahoud, and Hisham Cholakkal. 2024. Paris3d: Reasoning- based 3d part segmentation using large multimodal model. InEuropean Conference on Computer Vision. Springer, 466–482

  7. [15]

    Bernhard Kerbl, Georgios Kopanas, Thomas Leimkühler, and George Drettakis

  8. [16]

    Justin* Kerr, Chung Min* Kim, Ken Goldberg, Angjoo Kanazawa, and Matthew Tancik. 2023. LERF: Language Embedded Radiance Fields. InInternational Con- ference on Computer Vision (ICCV)

  9. [17]

    Prannay Khosla, Piotr Teterwak, Chen Wang, Aaron Sarna, Yonglong Tian, Phillip Isola, Aaron Maschinot, Ce Liu, and Dilip Krishnan. 2020. Supervised contrastive learning.Advances in neural information processing systems33 (2020), 18661– 18673

  10. [18]

    Berg, and Wan-Yen Lo

    Alexander Kirillov, Eric Mintun, Nikhila Ravi, Hanzi Mao, Chloe Rolland, Laura Gustafson, Tete Xiao, Spencer Whitehead, Alexander C. Berg, and Wan-Yen Lo

  11. [19]

    Xin Lai, Zhuotao Tian, Yukang Chen, Yanwei Li, Yuhui Yuan, Shu Liu, and Jiaya Jia. 2024. Lisa: Reasoning segmentation via large language model. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition. 9579–9589

  12. [20]

    Haijie Li, Yanmin Wu, Jiarui Meng, Qiankun Gao, Zhiyao Zhang, Ronggang Wang, and Jian Zhang. 2025. Instancegaussian: Appearance-semantic joint gaussian representation for 3D instance-level perception. InProceedings of the Computer Vision and Pattern Recognition Conference. 14078–14088

  13. [21]

    Tao Lu, Mulin Yu, Linning Xu, Yuanbo Xiangli, Limin Wang, Dahua Lin, and Bo Dai. 2024. Scaffold-gs: Structured 3D gaussians for view-adaptive rendering. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 20654–20664

  14. [22]

    InProceedings of the IEEE/CVF international conference on computer vision

    Segment anything. InProceedings of the IEEE/CVF international conference on computer vision. 4015–4026

  15. [23]

    Juliette Marrie, Romain Menegaux, Michael Arbel, Diane Larlus, and Julien Mairal. 2025. LUDVIG: Learning-Free Uplifting of 2D Visual Features to Gaussian Splatting Scenes. InProceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)

  16. [24]

    Srinivasan, Matthew Tancik, Jonathan T

    Ben Mildenhall, Pratul P. Srinivasan, Matthew Tancik, Jonathan T. Barron, Ravi Ramamoorthi, and Ren Ng. 2022. NeRF: Representing scenes as neural radiance fields for view synthesis.Commun. ACM65, 1 (2022), 99–106

  17. [25]

    Ashkan Mirzaei, Tristan Aumentado-Armstrong, Konstantinos G Derpanis, Jonathan Kelly, Marcus A Brubaker, Igor Gilitschenski, and Alex Levinshtein

  18. [26]

    Xiaojian Ma, Silong Yong, Zilong Zheng, Qing Li, Yitao Liang, Song-Chun Zhu, and Siyuan Huang. 2022. Sqa3d: Situated question answering in 3d scenes.arXiv preprint arXiv:2210.07474(2022)

  19. [27]

    Minghan Qin, Wanhua Li, Jiawei Zhou, Haoqian Wang, and Hanspeter Pfister

  20. [28]

    Yansong Qu, Shaohui Dai, Xinyang Li, Jianghang Lin, Liujuan Cao, Shengchuan Zhang, and Rongrong Ji. 2024. GOI: Find 3D gaussians of interest with an optimizable open-vocabulary semantic-space hyperplane. InProceedings of the 32nd ACM International Conference on Multimedia. ACM

  21. [29]

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, and Jack Clark

  22. [30]

    InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Spin-nerf: Multiview segmentation and perceptual inpainting with neural radiance fields. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 20669–20679

  23. [31]

    Qucheng Peng, Benjamin Planche, Zhongpai Gao, Meng Zheng, Anwesa Choud- huri, Terrence Chen, Chen Chen, and Ziyan Wu. 2024. 3d vision-language gaussian splatting.arXiv preprint arXiv:2410.07577(2024)

  24. [32]

    Changyue Shi, Minghao Chen, Yiping Mao, Chuxiao Yang, Xinyuan Hu, Jiajun Ding, and Zhou Yu. 2025. REALM: An MLLM-Agent Framework for Open World 3D Reasoning Segmentation and Editing on Gaussian Splatting.arXiv preprint arXiv:2510.16410(2025)

  25. [33]

    Jin-Chuan Shi, Miao Wang, Hao-Bin Duan, and Shao-Hua Guan. 2024. Language embedded 3D gaussians for open-vocabulary scene understanding. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 5333– 5343

  26. [34]

    Junchi Wang and Lei Ke. 2024. Llm-seg: Bridging image segmentation and large language model reasoning. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 1765–1774

  27. [35]

    Zeming Wei, Junyi Lin, Yang Liu, Weixing Chen, Jingzhou Luo, Guanbin Li, and Liang Lin. 2025. 3daffordsplat: Efficient affordance reasoning with 3d gaussians. In Proceedings of the 33rd ACM International Conference on Multimedia. 2821–2830

  28. [36]

    Yanmin Wu, Jiarui Meng, Haijie Li, Chenming Wu, Yahao Shi, Xinhua Cheng, Chen Zhao, Haocheng Feng, Errui Ding, Jingdong Wang, et al. 2024. Opengaus- sian: Towards point-level 3d gaussian-based open vocabulary understanding. Advances in Neural Information Processing Systems37 (...

  29. [37]

    Tianhe Ren, Shilong Liu, Ailing Zeng, Jing Lin, Kunchang Li, He Cao, Jiayu Chen, Xinyu Huang, Yukang Chen, Feng Yan, et al. 2024. Grounded sam: Assembling open-world models for diverse visual tasks.arXiv preprint arXiv:2401.14159 (2024)

  30. [38]

    Zhongwei Ren, Zhicheng Huang, Yunchao Wei, Yao Zhao, Dongmei Fu, Jiashi Feng, and Xiaojie Jin. 2024. Pixellm: Pixel reasoning with large multimodal model. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 26374–26383

  31. [39]

    Mingqiao Ye, Martin Danelljan, Fisher Yu, and Lei Ke. 2024. Gaussian grouping: Segment and edit anything in 3D scenes. InEuropean conference on computer vision. Springer, 162–179

  32. [40]

    Ni, and Heung-Yeung Shum

    Hao Zhang, Feng Li, Shilong Liu, Lei Zhang, Hang Su, Jun Zhu, Lionel M. Ni, and Heung-Yeung Shum. 2022. DINO: DETR with improved DeNoising anchor boxes for end-to-end object detection. arXiv preprint arXiv:2203.03605

  33. [41]

    Shijie Zhou, Haoran Chang, Sicheng Jiang, Zhiwen Fan, Zehao Zhu, Dejia Xu, Pradyumna Chari, Suya You, Zhangyang Wang, and Achuta Kadambi. 2024. Feature 3Dgs: Supercharging 3d gaussian splatting to enable distilled feature fields. InProceedings of the IEEE/CVF Conference on Com...

  34. [44]

    Cilin Yan, Haochen Wang, Shilin Yan, Xiaolong Jiang, Yao Hu, Guoliang Kang, Weidi Xie, and Efstratios Gavves. 2024. Visa: Reasoning video object segmentation via large language models. InEuropean Conference on Computer Vision. Springer, 98–115

  35. [45]

    Yuqi Yang, Peng-Tao Jiang, Jing Wang, Hao Zhang, Kai Zhao, Jinwei Chen, and Bo Li. 2024. Empowering segmentation ability to multi-modal large language models.arXiv preprint arXiv:2403.14141(2024)

  36. [2021]

    In International conference on machine learning

    Learning transferable visual models from natural language supervision. In International conference on machine learning. PmLR, 8748–8763

  37. [2023]

    Graph.42, 4 (2023), 139–1

    3D gaussian splatting for real-time radiance field rendering.ACM Trans. Graph.42, 4 (2023), 139–1

  38. [2024]

    InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Langsplat: 3D language gaussian splatting. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 20051–20060

  39. [2025]

    In2025 International Conference on 3D Vision (3DV)

    Reason3d: Searching and reasoning 3d segmentation via large language model. In2025 International Conference on 3D Vision (3DV). IEEE, 1177–1186

  40. [2026]

    arXiv preprint arXiv:2607.18801(2026)

    ZeroSplat: Generalized Referring Segmentation in 3D Gaussian Splatting. arXiv preprint arXiv:2607.18801(2026)

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.