Pith. sign in

REVIEW 3 major objections 5 minor 79 references

SpatioLM: Towards General Physical Spatial Intelligence in Vision-Language Models

T0 review · 3 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read SpatioLM claims a 0.3B add-on module can give a frozen VLM state-of-the-art spatial reasoning without 3D inputs, reporting 71.6 on VSI-Bench.

desk verdict Solid spatial-VLM systems result; headline benchmark number rests on an unverified train/eval overlap claim. read the letter →

arxiv 2608.01899 v1 pith:XEE6XPA2 submitted 2026-08-03 cs.CV cs.CLcs.LG

Jing Wu , Jianhua Wu , Jiayi Guan , Jiahong Chen , Jinghui Lu , Hangjun Ye , Bingzhao Gao , Long Chen This is my paper · ORCID
classification cs.CVcs.CLcs.LG
keywords spatialintelligencevision-languagemodelsmetricdepthestimationunderstandingparameter-efficientsidemodulepseudo-depthsupervisionembodiedmanipulation
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

The paper tries to establish that a frozen vision-language model (VLM) can acquire strong spatial perception and understanding by adding a small, plug-and-play side module, with no 3D sensor inputs, no external spatial encoder, and no retraining of the backbone. If true, this resolves a practical dilemma: previous spatial-VLM methods either depend on depth maps, point clouds, or tool APIs at inference, or fine-tune the backbone and lose general-purpose ability. SpatioLM reports the first average score above 70 on the VSI-Bench spatial-understanding benchmark (71.6), top metric-depth results (83.5 on single-image, 69.0 on multi-image), and 83.8 on the relative-depth DA-2K benchmark, while keeping degradation on four general benchmarks within roughly 4%. The same side module transfers to embodied manipulation, raising the LIBERO robot-manipulation benchmark success rate from 79.2% to 91.0% in a discrete-action setting.

What carries the argument

The load-bearing mechanism is the Spatio-Vision Module (SV-Module): about 0.3B parameters of stacked Spatio-Vision Blocks placed alongside a frozen VLM, with each block alternating one frame-attention pass (intra-frame geometry) and one global-attention pass (cross-frame consistency), a pattern adapted from multi-view geometry transformers. Its outputs are fused into frozen language-block hidden states through zero-initialized projections, so the pretrained distribution is perturbed only gradually. The module is steered by three supervisions: a language-modeling loss, a Gram-matrix token-distillation loss toward a pretrained depth teacher, and dense geometric losses on pseudo depth and camera ray maps from a dual Dense Prediction Transformer (DPT) head. The DPT head is discarded at inference, and the whole module is a side path that can be toggled on or off.

What would settle it

Check the scene identifiers: list every scene and video source in the VSI-590K, ScanQA, and SQA3D training splits and intersect them with the scene identifiers of VSI-Bench, ScanQA val, and SQA3D test; any shared scene invalidates the headline 71.6, and re-running the evaluation with all overlapping scenes removed gives the honest number.

Watch

Extended reading notes

Core claim

SpatioLM's central claim is that the geometric knowledge needed for spatial reasoning already exists, latent, in a pretrained VLM's intermediate visual tokens, and can be elicited on demand without changing the VLM. The authors freeze both the vision encoder and the language model, and train only a Spatio-Vision Module built from alternating frame-level and global attention blocks plus zero-initialized projection layers that inject geometry-aware features into selected frozen language-model layers. During training, a dual DPT head predicts pseudo depth maps and camera ray maps, and a token-distillation loss aligns the module's features with a pretrained 3D foundation model; at inference the head is removed, the module runs once during prefilling, and ordinary text decoding follows. On this recipe the paper reports state-of-the-art numbers across spatial perception and understanding benchmarks, including an average of 71.6 on VSI-Bench, the first reported result above 70, while the frozen backbone limits general-capability degradation to about 3-4%.

Load-bearing premise

The central numbers depend on the claim that no scene appears in both the training corpus and the VSI-Bench, ScanQA-val, and SQA3D-test evaluation sets; if any source scene leaks across the split, the reported gains could come from memorization rather than spatial reasoning.

Editorial extensions

If this is right

  • A frozen VLM can gain metric depth, relative depth, and spatial question-answering ability by training only a roughly 0.3B side module, so spatial enhancement no longer forces full-model fine-tuning.
  • Because the VLM backbone stays frozen, the same base model can serve both ordinary and spatially augmented modes, with the SV-Module adding no per-token decoding overhead after a single prefill pass.
  • If the reported results replicate, SpatioLM is the strongest reported RGB-only spatial VLM without external geometry at inference, on both perception and understanding benchmarks.
  • The same side module transfers to embodied manipulation, improving LIBERO average success from 79.2% to 91.0% under discrete actions and from 95.3% to 96.3% under continuous actions on a 2B base model.

Reading between the lines

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

  • A natural next stress test is to detach the SV-Module from its training teacher: if the module is truly eliciting latent geometry rather than memorizing teacher outputs, performance should survive on scenes and depth distributions the teacher never saw.
  • Because the paper itself reports motion-heavy MVBench subtasks as the main casualty, supplementing the geometric losses with an explicit temporal-dynamics signal such as optical flow is a direct, testable extension that could recover action counting and moving-direction accuracy while keeping the spatial gains.
  • The scene-disjoint split is the crux to check: recomputing VSI-Bench, ScanQA, and SQA3D numbers after filtering any scene overlap between training and evaluation would settle whether the 71.6 reflects spatial reasoning or memorization.
  • If the side-module recipe generalizes across base VLMs, it offers a cheap way to retrofit existing deployments with spatial ability without re-certifying the whole model.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. This paper proposes SpatioLM, a parameter-efficient framework that augments a frozen vision-language model with a plug-and-play Spatio-Vision Module (SV-Module) to improve spatial perception and understanding without extra 3D priors or third-party spatial encoders at inference. The module is trained with pseudo depth and camera-ray supervision, a distillation loss from a 3D foundation model, and a language modeling loss, while the VLM backbone remains frozen. The authors report state-of-the-art results on VSI-Bench (Avg 71.6, described as the first to exceed 70), ScanQA, SQA3D, and their self-constructed metric-depth benchmarks (MD-S, MD-M, DR), competitive results on DA-2K, limited degradation on general-purpose benchmarks, and improved performance on LIBERO manipulation tasks when adapted to a vision-language-action model.

Significance. The paper has several genuine strengths. The central design is clean: a frozen backbone with a compact side module (about 0.3B trainable parameters) that can be activated or deactivated at inference, and the ablations cover the main components (SV-Module, Vision Token Supervision, Dense Geometric Supervision), loss-weight sensitivity, attention mechanism, and number of blocks. The public-benchmark results on DA-2K, ScanQA, SQA3D, and LIBERO are external and not produced by the authors, which reduces the risk of circular evaluation. The paper also explicitly states limitations, including pseudo-label noise, a motion-dynamics trade-off, and safety concerns, which is commendable. If the data-overlap concern is resolved, the reported VSI-Bench result would be a notable advance in spatial VLM capabilities without external geometric inputs. The main uncertainty is whether the training and evaluation corpora are truly scene-disjoint; this is an unverified condition rather than an observed failure, but it is load-bearing for the headline claims.

major comments (3)
  1. [Sec. 4.1, App. B.1] The assertion that 'all training samples are strictly sourced from the official training splits' is not sufficient to rule out train/evaluation overlap for the central claims. VSI-590K is a re-packaged multi-source corpus rather than a dataset with a canonical official split, and the self-constructed MD-S, MD-M, and DR benchmarks use sources (e.g., Waymo, ScanNet, KITTI) that also appear in the training corpus. No scene-level or frame-level overlap audit is reported. Because the headline result (71.6 on VSI-Bench, 'the first to exceed 70') is the most exposed to memorization if any training video or scene also appears in evaluation, please provide a quantitative overlap analysis (e.g., scene-ID matching or frame hashing) across all training and evaluation sets, and report the fraction of overlapping samples if any are found.
  2. [Sec. 4.2.1, App. B.2] Three of the four perception benchmarks (MD-S, MD-M, DR) are constructed by the authors, yet the paper reports no error bars, confidence intervals, or statistical significance tests. For close comparisons, such as the DA-2K scores of 83.8 for both SpatioLM and VST-7B-SFT, the absence of variance makes the claimed superiority difficult to assess. In addition, the benchmark construction is described at a high level (e.g., 'randomly select 10 spatial locations' and '500 evaluation samples per dataset') without releasing the exact evaluation files. Please release the constructed benchmarks and report standard errors or repeated-run variability so that the perception results are independently verifiable.
  3. [Sec. 4.3, Fig. 6] The comparison of general-capability degradation against Spatial-MLLM is not controlled. Figure 6 compares the relative drops of SpatioLM built on SenseNovaSI-8B or InternVL3.5-8B with the relative drop of Spatial-MLLM built on Qwen2.5-VL-3B; because the base models differ in size and architecture, the large relative drop of Spatial-MLLM cannot be attributed solely to the added spatial module. The own-base comparisons in Table 14 (drops of about 3-4%) are more meaningful and do support the claim that SpatioLM mildly degrades general capability, but the comparative statement against Spatial-MLLM should either be re-run on the same base models or softened to avoid an apples-to-oranges comparison.
minor comments (5)
  1. [Tables 10-11] There are typographical inconsistencies in model names, e.g., 'LLaV A-NeXT' instead of 'LLaVA-NeXT', and 'InternVL3.5' versus 'InternVL3.5-VL' is used inconsistently throughout the tables and text.
  2. [Eq. (3)] The joint distribution in Eq. (3) factorizes as P(Y_g | X_v, X_t) followed by text generation conditioned on Y_g, but the paper later describes the auxiliary geometric outputs as supervision rather than as inputs to generation. Please clarify whether Y_g is used as a conditioning variable in the language modeling loss or is only a training-side auxiliary output.
  3. [Sec. 3.2] The choice of the intermediate ViT layer is described as 'guided by empirical analysis' and 'layers around 16/24,' but no ablation or validation-set result is shown for this selection. Please specify the exact layer index and provide the evidence for this choice.
  4. [App. B.1] The citation for VSI-590K is given as (Yang et al., 2025c), but the associated reference is the Cambrian-S paper; please ensure that the attribution is correct and that the reader can locate the original VSI-590K source.
  5. [Table 4] The meaning of the 'P.T.' (pretraining on large-scale robot manipulation data) column is not defined in the table caption or the text. Since the paper emphasizes that SpatioLM does not use such pretraining, please define the column and clarify which baselines use it.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity by construction; the core results rest on external benchmarks and an external teacher, with the main residual risk being unverified train/eval split disjointness rather than circularity.

full rationale

The derivation chain is open-loop. The Spatio-Vision Module is trained with three external supervision signals: a language-modeling loss on VQA corpora (Eq. 7), a distillation loss toward Depth Anything V3 features (Eq. 8), and dense depth and camera-ray pseudo-labels from Depth Anything V3 (Eq. 9); none of these targets is produced by the model itself or defined as a function of the reported benchmark scores. Evaluation is performed on external benchmarks (VSI-Bench, ScanQA, SQA3D, DA-2K) plus the self-built MD-S, MD-M, and DR benchmarks, and Sec. 4.1 states that training samples are strictly sourced from official training splits with validation and test splits reserved for evaluation. The headline VSI-Bench result of 71.6 is not derived by construction from the training objective, and no equation in Sec. 3 equates a predicted quantity to a fitted input. The self-built MD-S, MD-M, and DR benchmarks draw on datasets such as Waymo and ScanNet that also appear in the training corpus, so the split-disjointness assertion is a load-bearing empirical condition that should be audited; however, absent evidence of actual overlap, this is a data-hygiene risk rather than circularity. Citations to SenseNovaSI, Depth Anything V3, VGGT, and Bolya et al. are used as external base models, teachers, or architectural inspirations, not as self-justifying authorities, and no uniqueness theorem is invoked. The acknowledged pseudo-label dependence in Sec. 5 is a limitation on ceiling accuracy, but it is an external-teacher dependency, not a self-referential loop.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The ledger shows the paper's contribution is conditional on two upstream inputs: a third-party depth teacher (Depth Anything V3) providing all dense geometric labels, and an asserted but unproven scene-level separation between training and evaluation data. The SV-Module adds roughly 0.3B trainable parameters and three tuned hyperparameters, with no newly postulated physical entities.

free parameters (3)
  • Loss weights (alpha, beta, gamma) = 0.6, 0.2, 0.2
    Chosen empirically in Sec. 3.3, shown to be insensitive in Table 6; still a free choice affecting the objective.
  • Number of SV-Blocks = 6
    Selected in Sec. 4.5 (Table 8) based on performance-efficiency trade-off; 4 blocks underperform, 8-12 saturate.
  • Intermediate ViT layer index = 16 out of 24
    Selected in Sec. 3.2 based on 'empirical analysis' and Bolya et al. 2026; drives which tokens feed the SV-Module.
assumptions (4)
  • domain assumption Intermediate ViT layers (around 16/24) carry stronger spatial information than the final layer.
    Invoked in Sec. 3.2 with citation to Bolya et al. 2026; load-bearing because the SV-Module receives only these tokens.
  • domain assumption Depth Anything V3 pseudo-labels for depth and camera ray maps are accurate enough to supervise physically coherent representations.
    Sec. 3.3 and Limitations admit pseudo-label noise imposes an upper bound; all dense supervision derives from this external teacher.
  • domain assumption Training and evaluation splits are scene-disjoint across all sourced datasets and benchmarks, especially VSI-590K versus VSI-Bench.
    Asserted in Sec. 4.1 and App. B.1 without scene-level overlap analysis; if false, headline scores are inflated.
  • domain assumption Alternating frame/global attention in SV-Blocks captures both intra-frame and inter-frame geometric consistency without harming the frozen LLM.
    Architectural assumption validated only by the Table 7 ablation; no formal or analytical support.

how reviews work

0 comments
Cite this review

Pith. "Pith review of SpatioLM: Towards General Physical Spatial Intelligence in Vision-Language Models." pith.science (2026). https://pith.science/paper/XEE6XPA2

@misc{pith2026260801899,
  author       = {Pith},
  title        = {Pith review of: SpatioLM: Towards General Physical Spatial Intelligence in Vision-Language Models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/XEE6XPA2}},
  note         = {Machine review of arXiv:2608.01899}
}
read the original abstract

Vision-Language Models (VLMs) perform well on commonsense reasoning tasks but struggle with visual spatial reasoning. Most existing solutions introduce extra 3D prior inputs or external spatial encoders, which increase complexity and degrade the underlying VLMs' general-purpose capabilities after spatial fine-tuning. To this end, we propose a parameter-efficient \textit{\textbf{Spatio}-vision \textbf{L}anguage \textbf{M}odels (SpatioLM)}, that enhances spatial intelligence without extra 3D prior inputs or third-party spatial encoders. Concretely, we design a plug-and-play and non-invasive spatio-vision module that elicits the spatial knowledge inherent in VLMs. Furthermore, we innovatively leverage pseudo depth and camera information as supervision to guide the model in learning physically coherent representations. Extensive experiments show that SpatioLM achieves significant improvements in diverse tasks, including spatial perception and understanding while effectively limiting the degradation of general capabilities. Notably, the model achieves an impressive score of 71.6 on the VSI-Bench (the first model to surpass 70). In addition, it attains competitive performance when transferred to embodied manipulation tasks. Code is available at \href{https://github.com/xiaomi-research/spatio-lm}{\faGithub~spatio-lm}.

Figures

Figures reproduced from arXiv: 2608.01899 by the authors.

Figure 1
Figure 1. We propose SpatioLM, a parameter-efficient framework that improves spatial intelligence in VLMs without extra 3D prior inputs or external spatial encoders. SpatioLM achieves SOTA performance on both spatial perception and understanding tasks, and can be effectively adapted to embodied manipulation tasks under both discrete and continuous action settings. Abstract Vision-Language Models (VLMs) perform well on commons… view at source ↗
Figure 2
Figure 2. (a) Methods that explicitly leverage 3D priors, such as depth maps, point clouds, or camera parameters. (b) Methods that introduce an additional spatial encoder to inject 3D information. (c) Ours: a purely 2D vision-language framework that elicits 3D geometric structure from pretrained VLM visual tokens, without relying on explicit 3D inputs. ing capabilities. A third prominent paradigm addresses 3D reasoning throug… view at source ↗
Figure 3
Figure 3. Overview of SpatioLM. SpatioLM augments a frozen VLM with a plug-and-play Spatio-Vision Module. The Spatio-Vision Module elicits geometry-aware features from visual tokens and injects them into language blocks via zero-initialized projections, enabling visual spatial reasoning while preserving VLM’s general-purpose ability. Training incorporates auxiliary pseudo depth and camera supervision, while inference requires… view at source ↗
Figures from the paper (10 more)
Figure 4
Figure 4. Figure 4: Qualitative depth-tasks on the MD-S, MD￾M, DA-2K, and DR benchmarks [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 5
Figure 5. Figure 5: Qualitative tasks on VSI-Bench [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 6
Figure 6. Figure 6: Performance drop on general-purpose capabilities. Results. We present the quantitative results on Tab. 2 and Tab. 3; SpatioLMSenseNovaSI-8B achieves the best per￾formance across all three benchmarks. On VSI-Bench, it ranks 1st with Avg = 71.6, the first to exceed 70, a…
Figure 7
Figure 7. Figure 7: Qualitative manipulation-tasks on the LIBERO benchmark [PITH_FULL_IMAGE:figures/full_fig_p008_7.png]
Figure 8
Figure 8. Figure 8: Visualization samples of our spatial perception benchmarks. The first row shows ground-truth single-image metric depth (MD-S), and the second row shows multi-image metric depth estimation (MD-M). and reason about consistent metric depth across viewpoints, constituting …
Figure 9
Figure 9. Figure 9: Visualization samples of spatial perception benchmarks. The first row shows relative depth estimation (DA-2K), and the second row shows depth-related multi-tasks (DR). Spatial Understanding Data. For high-level spatial understanding, we leverage large-scale 3D visual q…
Figure 10
Figure 10. Figure 10: Spatial perception task visualization on MD-S and DA-2K benchmarks. 24 [PITH_FULL_IMAGE:figures/full_fig_p024_10.png]
Figure 11
Figure 11. Figure 11: Spatial perception task visualization on MD-M benchmark. 25 [PITH_FULL_IMAGE:figures/full_fig_p025_11.png]
Figure 12
Figure 12. Figure 12: Spatial perception task visualization on DR benchmark. 26 [PITH_FULL_IMAGE:figures/full_fig_p026_12.png]
Figure 13
Figure 13. Figure 13: Spatial understanding task visualization on VSI-Bench. 27 [PITH_FULL_IMAGE:figures/full_fig_p027_13.png]

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

79 extracted references · 8 canonical work pages

  1. [1]

    International conference on machine learning , pages=

    Learning transferable visual models from natural language supervision , author=. International conference on machine learning , pages=. 2021 , organization=

    2021

  2. [2]

    Advances in neural information processing systems , volume=

    Visual instruction tuning , author=. Advances in neural information processing systems , volume=

  3. [3]

    IEEE transactions on pattern analysis and machine intelligence , volume=

    Vision-language models for vision tasks: A survey , author=. IEEE transactions on pattern analysis and machine intelligence , volume=. 2024 , publisher=

    2024

  4. [4]

    arXiv preprint arXiv:2503.01773 , year=

    Why is spatial reasoning hard for vlms? an attention mechanism perspective on focus areas , author=. arXiv preprint arXiv:2503.01773 , year=

    arXiv

  5. [5]

    arXiv preprint arXiv:2503.17349 , year=

    Beyond semantics: Rediscovering spatial awareness in vision-language models , author=. arXiv preprint arXiv:2503.17349 , year=

    arXiv

  6. [6]

    arXiv preprint arXiv:2509.18905 , year=

    How far are vlms from visual spatial intelligence? a benchmark-driven perspective , author=. arXiv preprint arXiv:2509.18905 , year=

    arXiv

  7. [7]

    arXiv preprint arXiv:2509.09332 , year=

    Omnieva: Embodied versatile planner via task-adaptive 3d-grounded and embodiment-aware reasoning , author=. arXiv preprint arXiv:2509.09332 , year=

    work page arXiv

  8. [8]

    arXiv preprint arXiv:2506.01946 , year=

    MLLMs Need 3D-Aware Representation Supervision for Scene Understanding , author=. arXiv preprint arXiv:2506.01946 , year=

    arXiv

Show all 79 references
  1. [9]

    arXiv preprint arXiv:2506.04308 , year=

    RoboRefer: Towards Spatial Referring with Reasoning in Vision-Language Models for Robotics , author=. arXiv preprint arXiv:2506.04308 , year=

  2. [10]

    Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

    Mm-spatial: Exploring 3d spatial understanding in multimodal llms , author=. Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

  3. [11]

    arXiv preprint arXiv:2510.13800 , year=

    Reasoning in Space via Grounding in the World , author=. arXiv preprint arXiv:2510.13800 , year=

  4. [12]

    Advances in Neural Information Processing Systems , volume=

    Spatialrgpt: Grounded spatial reasoning in vision-language models , author=. Advances in Neural Information Processing Systems , volume=

  5. [13]

    arXiv preprint arXiv:2505.12448 , year=

    SSR: Enhancing Depth Perception in Vision-Language Models via Rationale-Guided Spatial Reasoning , author=. arXiv preprint arXiv:2505.12448 , year=

  6. [14]

    arXiv preprint arXiv:2501.15830 , year=

    Spatialvla: Exploring spatial representations for visual-language-action model , author=. arXiv preprint arXiv:2501.15830 , year=

    arXiv

  7. [15]

    Proceedings of the Computer Vision and Pattern Recognition Conference , pages=

    Spatialllm: A compound 3d-informed design towards spatially-intelligent large multimodal models , author=. Proceedings of the Computer Vision and Pattern Recognition Conference , pages=

  8. [16]

    arXiv preprint arXiv:2505.20279 , year=

    VLM-3R: Vision-Language Models Augmented with Instruction-Aligned 3D Reconstruction , author=. arXiv preprint arXiv:2505.20279 , year=

    arXiv

  9. [17]

    arXiv preprint arXiv:2505.24625 , year=

    Learning from Videos for 3D World: Enhancing MLLMs with 3D Vision Geometry Priors , author=. arXiv preprint arXiv:2505.24625 , year=

  10. [18]

    arXiv preprint arXiv:2505.23747 , year=

    Spatial-mllm: Boosting mllm capabilities in visual-based spatial intelligence , author=. arXiv preprint arXiv:2505.23747 , year=

    arXiv

  11. [19]

    arXiv preprint arXiv:2510.13375 , year=

    DepthVLA: Enhancing Vision-Language-Action Models with Depth-Aware Spatial Reasoning , author=. arXiv preprint arXiv:2510.13375 , year=

  12. [20]

    arXiv preprint arXiv:2509.25413 , year=

    Depthlm: Metric depth from vision language models , author=. arXiv preprint arXiv:2509.25413 , year=

  13. [21]

    arXiv preprint arXiv:2511.04670 , year=

    Cambrian-s: Towards spatial supersensing in video , author=. arXiv preprint arXiv:2511.04670 , year=

    arXiv

  14. [22]

    arXiv preprint arXiv:2410.02073 , year=

    Depth pro: Sharp monocular metric depth in less than a second , author=. arXiv preprint arXiv:2410.02073 , year=

    arXiv

  15. [23]

    IEEE Transactions on Pattern Analysis and Machine Intelligence , year=

    Metric3d v2: A versatile monocular geometric foundation model for zero-shot metric depth and surface normal estimation , author=. IEEE Transactions on Pattern Analysis and Machine Intelligence , year=

  16. [24]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    Spatialvlm: Endowing vision-language models with spatial reasoning capabilities , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  17. [25]

    2025 IEEE International Conference on Robotics and Automation (ICRA) , pages=

    Spatialbot: Precise spatial understanding with vision language models , author=. 2025 IEEE International Conference on Robotics and Automation (ICRA) , pages=. 2025 , organization=

    2025

  18. [26]

    arXiv preprint arXiv:2511.05491 , year=

    Visual spatial tuning , author=. arXiv preprint arXiv:2511.05491 , year=

  19. [27]

    arXiv preprint arXiv:2511.13719 , year=

    Scaling Spatial Intelligence with Multimodal Foundation Models , author=. arXiv preprint arXiv:2511.13719 , year=

  20. [28]

    arXiv preprint arXiv:2406.09246 , year=

    Openvla: An open-source vision-language-action model , author=. arXiv preprint arXiv:2406.09246 , year=

    arXiv

  21. [29]

    CoRR, abs/2410.24164, 2024

    0: A vision-language-action flow model for general robot control. CoRR, abs/2410.24164, 2024. doi: 10.48550 , author=. arXiv preprint ARXIV.2410.24164 , year=

    2024 arXiv

  22. [30]

    arXiv preprint arXiv:2504.16054 , year=

    0.5: a Vision-Language-Action Model with Open-World Generalization , author=. arXiv preprint arXiv:2504.16054 , year=

    arXiv

  23. [31]

    arXiv preprint arXiv:2503.14734 , year=

    Gr00t n1: An open foundation model for generalist humanoid robots , author=. arXiv preprint arXiv:2503.14734 , year=

    arXiv

  24. [32]

    arXiv preprint arXiv:2505.21906 , year=

    Vision-Language-Action Model with Open-World Embodied Reasoning from Pretrained Knowledge , author=. arXiv preprint arXiv:2505.21906 , year=

    arXiv

  25. [33]

    Proceedings of the Computer Vision and Pattern Recognition Conference , pages=

    Vggt: Visual geometry grounded transformer , author=. Proceedings of the Computer Vision and Pattern Recognition Conference , pages=

  26. [34]

    arXiv preprint arXiv:2511.10647 , year=

    Depth anything 3: Recovering the visual space from any views , author=. arXiv preprint arXiv:2511.10647 , year=

    arXiv

  27. [35]

    Proceedings of the Computer Vision and Pattern Recognition Conference , pages=

    Thinking in space: How multimodal large language models see, remember, and recall spaces , author=. Proceedings of the Computer Vision and Pattern Recognition Conference , pages=

  28. [36]

    arXiv preprint arXiv:2510.13054 , year=

    Vla-0: Building state-of-the-art vlas with zero modification , author=. arXiv preprint arXiv:2510.13054 , year=

  29. [37]

    arXiv preprint arXiv:2502.19645 , year=

    Fine-tuning vision-language-action models: Optimizing speed and success , author=. arXiv preprint arXiv:2502.19645 , year=

    arXiv

  30. [38]

    Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    Scan2cap: Context-aware dense captioning in rgb-d scans , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  31. [39]

    Proceedings of the Computer Vision and Pattern Recognition Conference , pages=

    Video-mme: The first-ever comprehensive evaluation benchmark of multi-modal llms in video analysis , author=. Proceedings of the Computer Vision and Pattern Recognition Conference , pages=

  32. [40]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    Mvbench: A comprehensive multi-modal video understanding benchmark , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  33. [41]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    Mmmu: A massive multi-discipline multimodal understanding and reasoning benchmark for expert agi , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  34. [42]

    arXiv preprint arXiv:2501.13826 , year=

    Video-mmmu: Evaluating knowledge acquisition from multi-discipline professional videos , author=. arXiv preprint arXiv:2501.13826 , year=

    arXiv

  35. [43]

    Advances in Neural Information Processing Systems , volume=

    Libero: Benchmarking knowledge transfer for lifelong robot learning , author=. Advances in Neural Information Processing Systems , volume=

  36. [44]

    arXiv preprint arXiv:2505.06111 , year=

    Univla: Learning to act anywhere with task-centric latent actions , author=. arXiv preprint arXiv:2505.06111 , year=

    arXiv

  37. [45]

    arXiv preprint arXiv:2506.01844 , year=

    Smolvla: A vision-language-action model for affordable and efficient robotics , author=. arXiv preprint arXiv:2506.01844 , year=

    arXiv

  38. [46]

    arXiv preprint arXiv:2508.18269 , year=

    Flowvla: Visual chain of thought-based motion reasoning for vision-language-action models , author=. arXiv preprint arXiv:2508.18269 , year=

  39. [47]

    arXiv preprint arXiv:2503.15558 , year=

    Cosmos-reason1: From physical common sense to embodied reasoning , author=. arXiv preprint arXiv:2503.15558 , year=

    arXiv

  40. [48]

    5-vl technical report , author=

    Qwen2. 5-vl technical report , author=. arXiv preprint arXiv:2502.13923 , year=

    arXiv

  41. [49]

    2025 , eprint=

    Qwen3-VL Technical Report , author=. 2025 , eprint=

    2025

  42. [50]

    arXiv preprint arXiv:2508.18265 , year=

    Internvl3.5: Advancing open-source multimodal models in versatility, reasoning, and efficiency , author=. arXiv preprint arXiv:2508.18265 , year=

    arXiv

  43. [51]

    LLaVA-NeXT: A Strong Zero-shot Video Understanding Model , url=

    Zhang, Yuanhan and Li, Bo and Liu, haotian and Lee, Yong jae and Gui, Liangke and Fu, Di and Feng, Jiashi and Liu, Ziwei and Li, Chunyuan , month=. LLaVA-NeXT: A Strong Zero-shot Video Understanding Model , url=

  44. [52]

    2025 , howpublished =

    Introducing GPT-5.2 , author =. 2025 , howpublished =

    2025

  45. [53]

    arXiv preprint arXiv:2505.07062 , year=

    Seed1.5-vl technical report , author=. arXiv preprint arXiv:2505.07062 , year=

    arXiv

  46. [54]

    arXiv preprint arXiv:2507.06261 , year=

    Gemini 2.5: Pushing the frontier with advanced reasoning, multimodality, long context, and next generation agentic capabilities , author=. arXiv preprint arXiv:2507.06261 , year=

    arXiv

  47. [55]

    arXiv preprint arXiv:2511.16518 , year=

    MiMo-Embodied: X-Embodied Foundation Model Technical Report , author=. arXiv preprint arXiv:2511.16518 , year=

    arXiv

  48. [56]

    2025 , eprint=

    OpenAI GPT-5 System Card , author=. 2025 , eprint=

    2025

  49. [57]

    Advances in Neural Information Processing Systems , volume=

    Depth anything v2 , author=. Advances in Neural Information Processing Systems , volume=

  50. [58]

    Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

    Scannet: Richly-annotated 3d reconstructions of indoor scenes , author=. Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

  51. [59]

    Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

    Scannet++: A high-fidelity dataset of 3d indoor scenes , author=. Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

  52. [60]

    arXiv preprint arXiv:2111.08897 , year=

    Arkitscenes: A diverse real-world dataset for 3d indoor scene understanding using mobile rgb-d data , author=. arXiv preprint arXiv:2111.08897 , year=

    arXiv

  53. [61]

    Proceedings of the IEEE/CVF international conference on computer vision , pages=

    Hypersim: A photorealistic synthetic dataset for holistic indoor scene understanding , author=. Proceedings of the IEEE/CVF international conference on computer vision , pages=

  54. [62]

    Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    Scalability in perception for autonomous driving: Waymo open dataset , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  55. [63]

    Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

    Deepmvs: Learning multi-view stereopsis , author=. Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

  56. [64]

    arXiv preprint arXiv:2001.10773 , year=

    Virtual kitti 2 , author=. arXiv preprint arXiv:2001.10773 , year=

    2001 arXiv

  57. [65]

    arXiv preprint arXiv:2210.07474 , year=

    Sqa3d: Situated question answering in 3d scenes , author=. arXiv preprint arXiv:2210.07474 , year=

    arXiv

  58. [66]

    proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    Scanqa: 3d question answering for spatial scene understanding , author=. proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  59. [67]

    Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

    Sun rgb-d: A rgb-d scene understanding benchmark suite , author=. Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

  60. [68]

    European conference on computer vision , pages=

    Indoor segmentation and support inference from rgbd images , author=. European conference on computer vision , pages=. 2012 , organization=

    2012

  61. [69]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    Neural rgb-d surface reconstruction , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  62. [70]

    The international journal of robotics research , volume=

    Vision meets robotics: The kitti dataset , author=. The international journal of robotics research , volume=. 2013 , publisher=

    2013

  63. [71]

    Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

    Scene coordinate regression forests for camera relocalization in RGB-D images , author=. Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

  64. [72]

    arXiv preprint arXiv:2410.21276 , year=

    Gpt-4o system card , author=. arXiv preprint arXiv:2410.21276 , year=

    arXiv

  65. [73]

    Advances in Neural Information Processing Systems , volume=

    Perception encoder: The best visual embeddings are not at the output of the network , author=. Advances in Neural Information Processing Systems , volume=

  66. [74]

    Advances in Neural Information Processing Systems , volume=

    Deepstack: Deeply stacking visual tokens is surprisingly simple and effective for lmms , author=. Advances in Neural Information Processing Systems , volume=

  67. [75]

    Proceedings of the IEEE/CVF international conference on computer vision , pages=

    Vipergpt: Visual inference via python execution for reasoning , author=. Proceedings of the IEEE/CVF international conference on computer vision , pages=

  68. [76]

    Proceedings of the Computer Vision and Pattern Recognition Conference , pages=

    Visual agentic ai for spatial reasoning with a dynamic api , author=. Proceedings of the Computer Vision and Pattern Recognition Conference , pages=

  69. [77]

    Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    Visual programming: Compositional visual reasoning without training , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  70. [78]

    Advances in Neural Information Processing Systems , volume=

    Grounded reinforcement learning for visual reasoning , author=. Advances in Neural Information Processing Systems , volume=

  71. [79]

    arXiv preprint arXiv:2512.08889 , year=

    No Labels, No Problem: Training Visual Reasoners with Multimodal Verifiers , author=. arXiv preprint arXiv:2512.08889 , year=

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.