Pith. sign in

REVIEW 3 major objections 4 minor 35 references

Privileged-evidence distillation can convert training-time 3D cues into RGB-only spatial reasoning, making deployable models stronger.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · deepseek-v4-flash

2026-08-04 00:34 UTC pith:T324FV4V

load-bearing objection Useful distillation framework for RGB-only 3D-VLMs, but the central attribution of student gains to privileged evidence is underdetermined without an RGB-only teacher control. the 3 major comments →

arxiv 2608.00110 v1 pith:T324FV4V submitted 2026-07-31 cs.CV

Distill What RGB Can Recover: Privileged 3D Evidence for RGB-Only Vision-Language Models

Yanbin Hu , Jin Cui , Jun Ye , Jiepeng Zhou , Jiangcheng Song , Boran Zhao , Pengju Ren This is my paper
classification cs.CV
keywords privileged-evidence distillationRGB-only 3D vision-language modelsevidence-sensitivity weightingspatial reasoning3D scene understandingknowledge distillationrecoverability decompositionhallucination reduction
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The paper tries to establish that a vision-language model that sees only RGB images at deployment can still acquire genuine 3D perception—depth ordering, object relations, spatial layout—if that evidence is used at training time to supervise a teacher, and the teacher's knowledge is selectively distilled into the RGB-only student. It argues that naive imitation of a 3D-enabled teacher is wasteful because some teacher outputs depend on evidence that RGB cannot recover; the correct strategy is to estimate each sample's dependence on 3D evidence and down-weight supervision on highly evidence-dependent samples. The paper reports an RGB-only student that outperforms a carefully matched RGB-only baseline on all 11 benchmark metrics across four 3D scene-understanding tasks, with the largest gains on answer generation and captioning. If true, this means depth sensors and 3D reconstructions are not needed at inference time to get substantially better spatial reasoning—just better training.

Core claim

The central claim is that a teacher's privileged 3D gain over a matched RGB-only baseline decomposes into a part RGB can recover and a residual gap, and evidence-sensitivity-guided distillation transfers the recoverable part. The paper's experiments show an RGB-only student beating the baseline on all 11 metrics—10.4 CIDEr on ScanQA and 19.1 CIDEr@0.5 on Scan2Cap—while receiving only RGB frames and a question at inference. Recovery varies by task and skill: 90.4% of the teacher gain on ScanQA but 16.0% on SQA3D, and metric-distance reasoning recovers only about 45%. Transferability is not a fixed property of evidence type; it depends on how much the teacher's target relies on evidence RGB ca

What carries the argument

The load-bearing mechanism is the Unified Evidence Token interface combined with evidence-sensitivity weighting. Depth, object, and relation inputs are encoded into a shared space, combined by a question-conditioned router, compressed into a fixed number of tokens by Q-Former-style cross-attention, and injected as small residual cross-attention updates into selected language layers, so the teacher stays close to the original RGB pathway. To decide what to transfer, the method corrupts evidence by sample-mismatched depth and shuffled object/relation features, measures the Jensen-Shannon divergence between teacher predictions under clean vs corrupted evidence, and down-weights distillation los

Load-bearing premise

The whole weighting scheme rests on the assumption that corrupting depth by mismatched samples and shuffling object/relation features disrupts only the semantic correspondence of the evidence, while preserving its scale and marginal statistics; if the corruption also disturbs the teacher in unintended ways, the sensitivity scores, the distillation weights, and the reported recovery decomposition would all be biased.

What would settle it

Take a held-out set of scenes with known ground-truth geometry and split it into samples where the answer is recoverable from monocular cues (e.g., relative depth of two visible objects) and samples where it is not (e.g., metric distance to an object fully occluded in the image). Compute the teacher's evidence-sensitivity weights on each subset. The paper's explanation predicts high sensitivity—and hence down-weighting—for the unrecoverable subset and low sensitivity for the recoverable subset. A failure to find this separation, or a re-training without sensitivity weighting that matches the s

Watch this falsifier. Get emailed when new claim-graph text bears on it.

Share X Bluesky LinkedIn Reddit HN

If this is right

  • If the claim holds, 3D-VLMs can be deployed with only RGB images and questions, avoiding depth sensors, point-cloud reconstruction, and 3D annotation at inference time.
  • Spatial reasoning gains become a training-time cost: any available 3D evidence—sensors, scans, or pseudo-depth—can be injected and selectively distilled without changing the deployable model.
  • The evidence-sensitivity weighting appears to suppress hallucination and affirmative-answer bias relative to uniform distillation, with 3D-POPE F1 up 8.7 points and Yes% down 24.8 points.
  • Per-skill recovery results imply that relative-depth and directional-relation skills are mostly RGB-recoverable, while metric-distance reasoning remains strongly dependent on explicit 3D evidence.
  • The teacher's final-layer injection ablation shows that evidence injection must stay off the final layer; injecting there can push the affirmative-answer rate to 100% on the diagnostic.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • A practical triage follows from the recoverability heatmap: invest in additional sensors or 3D annotation chiefly for skills with low recovery (metric distance, some counting), since RGB may never fully internalize them.
  • The corruption-based sensitivity estimator may transfer to other privileged modalities; if it does, its validity would be confirmed by showing the learned weights are stable across different corruption operators.
  • One can test the paper's explanation directly with synthetic scenes of known visibility: if sensitivity-weighted distillation down-weights samples whose target object is occluded and the student only improves on visible geometries, the recoverability picture is causal rather than correlational.
  • The matched-baseline design already controls optimization steps and capacity, but an extra RGB-only control trained with monocular augmentation (e.g., varying viewpoint videos) would say whether the gains are unique to teacher distillation or just to more effective training signal.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 4 minor

Summary. The paper proposes a privileged-evidence distillation framework for RGB-only 3D vision-language models. A teacher T is trained with depth, object, and relation evidence encoded through a unified evidence interface and injected into selected language layers via sparse residual cross-attention. An RGB-only student S is trained from the frozen teacher using logit distillation, structural (token-interaction) distillation, and evidence-sensitivity weighting that down-weights teacher targets highly dependent on privileged evidence. A matched RGB-only baseline B is trained independently with the same backbone, data, and optimization budget. The paper reports that S outperforms B on all 11 metrics across ScanQA, SQA3D, ScanRefer, and Scan2Cap (e.g., +10.4 ScanQA CIDEr, +19.1 Scan2Cap CIDEr@0.5), and defines a recoverability decomposition Δ_priv = Δ_trans + Δ_res, with transfer ratios R = Δ_trans/Δ_priv.

Significance. The matched baseline is a genuine strength: B and S share initialization, data, trainable-parameter budget, and number of optimization steps, so the comparison S versus B is far cleaner than typical unconstrained KD comparisons. The recoverability decomposition is a transparent and useful accounting identity, and the evidence-sensitivity weighting idea is interesting and plausibly valuable for avoiding imitation of RGB-unsupported teacher behavior. If the attribution to privileged evidence is substantiated, the paper would offer a practical deployment-oriented direction for RGB-only 3D reasoning. However, the central attribution claim is currently underdetermined: no strong RGB-only teacher control isolates privileged-evidence transfer from generic knowledge-distillation benefits, and all results are single-run point estimates without statistical reliability. The corruption operator used for sensitivity weighting and leave-one-out analysis is also asserted, not verified.

major comments (3)
  1. [RGB-Only Student Distillation / Eq. (18)] The main claim that S−B is a 'transferred privileged gain' requires a control that is absent. The comparison S (task loss + distillation from T) versus B (task loss only) cannot separate the effect of the teacher's privileged 3D evidence from the generic benefit of distilling from a stronger model. Table 4 shows that the teacher with corrupted evidence (97.2 ScanQA CIDEr) is close to B (96.8), which helps rule out the extra interface as the source of teacher strength, but it does not rule out generic KD: a student distilled from a strong RGB-only teacher (e.g., a teacher with more LoRA capacity, an ensemble, or longer training) might achieve a similar S−B gain without any privileged evidence. Please add such an RGB-only-teacher distillation control with the same student objective, or explicitly weaken the attribution language in the abstract, Eq. (18), and the recoverability ratio discus
  2. [Table 1 / Table 3 / Experimental Setup] All reported results are single-run point estimates with no error bars or significance tests. This matters for the 'all 11 metrics' claim and the recovery ratios: some differences are small in absolute terms, e.g., SQA3D EM S−B = 0.4 and ScanQA BLEU-4 S−B = 0.4 in Table 1; in Table 3 the sensitivity-weighting comparison is 104.8 vs 107.2 CIDEr and 48.9% vs 67.0% recovery. Without multiple seeds or paired bootstrap/statistical tests, the central claim that the student 'consistently outperforms' the baseline could be partly noise. Please report standard deviations over at least 3 runs, or provide paired tests across evaluation examples.
  3. [Eqs. (14)-(15) and Table 4] The corruption operator is load-bearing for the evidence-sensitivity weights w(x) and for the leave-one-evidence-out analysis in Table 4. The paper asserts that sample-mismatched depth and shuffled object/relation features 'approximately preserve feature scale and marginal statistics,' but no verification is provided. If the corruption disturbs teacher behavior beyond the intended counterfactual (e.g., by introducing out-of-distribution features), the sensitivity scores would be biased and the distillation weights miscomputed. Please provide diagnostics (e.g., feature mean/variance and teacher-logit statistics under clean vs corrupted evidence), and test sensitivity of the main results to an alternative corruption operator (e.g., moment-matched Gaussian noise or independent permutation variants).
minor comments (4)
  1. [Eq. (18)] The notation cM(T) is confusing; c appears to be a metric value, M a metric index, and the superscript notation is nonstandard. Please define the metric M and write M(T), M(S), M(B) or similar.
  2. [Table 4] The row 'Evidence off' is evaluated with the corruption operator applied, so it is not the same as an RGB-only teacher. Clarify this in the caption to avoid implying that the teacher is evaluated without evidence in the ordinary sense.
  3. [Figure 3] Recovery percentages in the lower heatmap are reported without denominators, sample sizes, or uncertainty. Cells with 100% recovery are particularly sensitive to near-zero denominators; please provide sample counts and standard errors, or mark cells with small sample sizes.
  4. [General] No code or data release is mentioned. Given the number of hyperparameters (κ, λ_kd, λ_str, Lv, Nu) and the complexity of the two-phase teacher training, releasing code would substantially improve reproducibility.

Circularity Check

0 steps flagged

No circularity: the student, teacher, and baseline numbers are measured outcomes, and the recoverability decomposition is an explicit algebraic identity rather than a hidden derivation.

full rationale

The paper's claimed chain is empirical, not tautological. The teacher T is trained with additional privileged evidence, the RGB-only baseline B with task supervision only, and the student S with task supervision plus logit/structural distillation weighted by evidence sensitivity. The reported student gains, e.g. S-B on ScanQA CIDEr and Scan2Cap CIDEr@0.5, are forward measurements, not quantities fitted from the teacher and then relabeled as predictions. The recoverability decomposition in Eqs. 18-19 is explicitly defined as the identity cM(T)-cM(B) = [cM(S)-cM(B)] + [cM(T)-cM(S)], so it cannot force any empirical outcome; it is arithmetic bookkeeping applied to measured values. The sensitivity weights w(x) are estimated from teacher self-dependence and used to modulate distillation, but the ground-truth loss remains unweighted, so the student's final performance is not determined by construction. The lack of a strong RGB-only teacher control is a genuine validity/confound concern (generic knowledge-distillation benefit is not isolated), and the claim that the corruption operator "approximately preserves feature scale and marginal statistics" is unverified, but both are correctness or assumption issues rather than circularity. There are no load-bearing self-citations and no imported uniqueness theorems. Therefore no significant circularity is present.

Axiom & Free-Parameter Ledger

4 free parameters · 4 axioms · 3 invented entities

The central claim rests on standard deep learning machinery, training-time privileged evidence (including ground-truth annotations), and an unverified corruption operator. The main free parameters are loss weights and architectural choices not tuned by rigorous sensitivity analysis.

free parameters (4)
  • κ = not reported
    Attenuation coefficient in w(x)=exp(-κ s(x)); chosen by hand, controls down-weighting strength.
  • λ_kd, λ_str = not reported
    Weights for logit and structural distillation losses in Eq. (16); chosen by hand, no sensitivity analysis.
  • Lv (virtual fusion slots) = not reported
    Number of slots in the question-conditioned router; architectural hyperparameter.
  • Nu (unified evidence tokens) = 80
    Number of evidence tokens; chosen for context length/performance tradeoff.
axioms (4)
  • standard math Transformer backbone with LoRA fine-tuning and Q-Former-style cross-attention are used as base machinery.
    Assumes standard deep learning operations and architectures perform as expected; no proof provided.
  • domain assumption Privileged evidence E = {depth D, object descriptors O, pairwise relations R} is available during training for all benchmarks.
    The method depends on training-time access to these signals, e.g., ground-truth object annotations for ScanRefer/Scan2Cap.
  • ad hoc to paper The corruption operator (sample-mismatched depth + shuffled object/relation features) isolates evidence dependence while preserving feature scale and marginals.
    Central to evidence-sensitivity weighting; validity is stated but not verified.
  • domain assumption RGB inputs contain sufficient monocular cues for a substantial portion of privileged gains to be recoverable.
    The whole approach assumes some transfer is possible; the paper's own Fig. 3 shows recovery varies by skill.
invented entities (3)
  • Unified Evidence Tokens no independent evidence
    purpose: Compress depth, object, and relation evidence into a fixed set of tokens injected into selected LLM layers.
    Internal representation; no falsifiable handle outside the benchmarks reported.
  • Question-conditioned router no independent evidence
    purpose: Predicts per-question fusion weights over the three evidence streams.
    Learned module; its outputs are observable only through ablation comparisons.
  • Evidence-sensitivity weight w(x) no independent evidence
    purpose: Down-weights distillation loss for samples whose teacher predictions are highly sensitive to corrupted evidence.
    Depends on the corruption operator; not independently validated.

pith-pipeline@v1.3.0-alltime-deepseek · 11993 in / 11474 out tokens · 109348 ms · 2026-08-04T00:34:34.260929+00:00 · methodology

0 comments
Cite this review

Pith. "Pith review of Distill What RGB Can Recover: Privileged 3D Evidence for RGB-Only Vision-Language Models." pith.science (2026). https://pith.science/paper/T324FV4V

@misc{pith2026260800110,
  author       = {Pith},
  title        = {Pith review of: Distill What RGB Can Recover: Privileged 3D Evidence for RGB-Only Vision-Language Models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/T324FV4V}},
  note         = {Machine review of arXiv:2608.00110}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

3D scene understanding requires reasoning about entity existence, spatial layout, and object relations, yet RGB images alone often provide insufficient 3D cues. Existing 3D-VLMs commonly rely on depth or 3D-position-aware inputs at inference time, introducing additional acquisition, reconstruction, or annotation costs that limit RGB-only deployment. We therefore study how training-time 3D evidence can be converted into spatial reasoning capabilities retained under RGB-only inference. We propose a privileged-evidence distillation framework that constructs a distillable teacher through a unified evidence interface and controlled residual injection, and transfers its knowledge to a deployable student receiving only RGB images and questions through logit and structured representation distillation. To avoid imitating teacher signals unsupported by RGB, we further introduce evidence-sensitivity-guided distillation, which uses corrupted evidence to identify highly evidence-dependent targets and down-weight their supervision. We also define a recoverability decomposition based on the matched baseline, teacher, and student, separating privileged gains into RGB-recoverable improvements and residual teacher advantages. Across four benchmarks, the teacher achieves the best result on 7 of 11 reported metrics among the compared methods. The RGB-only student outperforms its matched baseline on all 11 metrics, including gains of 10.4 ScanQA CIDEr and 19.1 Scan2Cap CIDEr@0.5, without additional inference-time inputs. These results validate the effectiveness of training-time privileged 3D evidence distillation for both teacher performance and deployable RGB-only spatial reasoning. Separately, our matched baseline-teacher-student analysis characterizes privileged-gain transfer across evidence types and spatial skills.

Figures

Figures reproduced from arXiv: 2608.00110 by Boran Zhao, Jiangcheng Song, Jiepeng Zhou, Jin Cui, Jun Ye, Pengju Ren, Yanbin Hu.

Figure 1
Figure 1. Figure 1: Conceptual overview of privileged-evidence dis￾tillation. The teacher (T) uses training-time depth, object, and relation evidence, while the student (S) learns trans￾ferable spatial reasoning from RGB only. Relative to the matched RGB-only baseline (B), the student recovers part of the privileged gain, while the remaining gap represents knowledge not transferred to RGB-only inference. scene geometry, objec… view at source ↗
Figure 2
Figure 2. Figure 2: Overview of our privileged-evidence distillation framework. During training, depth, object, and relation evidence is fused into Unified Evidence Tokens and injected into selected language layers to construct the privileged teacher T. The RGB-only student S learns from the teacher through logit and structural distillation while retaining the original RGB-only input pathway. At inference, all evidence module… view at source ↗
Figure 3
Figure 3. Figure 3: Evidence- and skill-wise privileged gain and re￾coverability.The upper heatmap reports the teacher improve￾ment over the RGB-only baseline, and the lower heatmap re￾ports the fraction of that improvement retained by the student. Rows correspond to depth, object, relation, and aggregate ev￾idence groups; columns correspond to spatial-skill subsets. Blank cells indicate teacher gains below ϵ. tance score; th… view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

35 extracted references · 7 linked inside Pith

  1. [1]

    Anderson, P.; Wu, Q.; Teney, D.; Bruce, J.; Johnson, M.; S \"u nderhauf, N.; Reid, I.; Gould, S.; and Van Den Hengel, A. 2018. Vision-and-language navigation: Interpreting visually-grounded navigation instructions in real environments. In Proceedings of the IEEE conference on computer vision and pattern recognition, 3674--3683

    2018

  2. [2]

    Azuma, D.; Miyanishi, T.; Kurita, S.; and Kawanabe, M. 2022. Scanqa: 3d question answering for spatial scene understanding. In proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 19129--19139

    2022

  3. [3]

    Z.; Chang, A

    Chen, D. Z.; Chang, A. X.; and Nie ner, M. 2020. Scanrefer: 3d object localization in rgb-d scans using natural language. In European conference on computer vision, 202--221. Springer

    2020

  4. [4]

    Chen, S.; Chen, X.; Zhang, C.; Li, M.; Yu, G.; Fei, H.; Zhu, H.; Fan, J.; and Chen, T. 2024. Ll3da: Visual interactive instruction tuning for omni-3d understanding reasoning and planning. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 26428--26438

    2024

  5. [5]

    Chen, S.; Zhu, H.; Chen, X.; Lei, Y.; Yu, G.; and Chen, T. 2023. End-to-end 3d dense captioning with vote2cap-detr. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 11124--11133

    2023

  6. [6]

    Chen, Z.; Gholami, A.; Nie ner, M.; and Chang, A. X. 2021. Scan2cap: Context-aware dense captioning in rgb-d scans. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 3193--3203

    2021

  7. [7]

    Chen, Z.; Zhang, X.; Xu, H.; Xie, J.; and Tu, Z. 2026. Cvp: Central-peripheral vision-inspired multimodal model for spatial reasoning. In Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision, 2295--2305

    2026

  8. [8]

    Cheng, A.-C.; Yin, H.; Fu, Y.; Guo, Q.; Yang, R.; Kautz, J.; Wang, X.; and Liu, S. 2024. Spatialrgpt: Grounded spatial reasoning in vision-language models. Advances in Neural Information Processing Systems, 37: 135062--135093

    2024

  9. [9]

    Ciernik, L.; Morik, M.; Thede, L.; Eyring, L.; Nakajima, S.; Akata, Z.; and Muttenthaler, L. 2026. Beyond the final layer: Attentive multilayer fusion for vision transformers. arXiv preprint arXiv:2601.09322

    Pith/arXiv arXiv 2026

  10. [10]

    Dai, R.; Das, S.; and Bremond, F. 2021. Learning an augmented rgb representation with cross-modal knowledge distillation for action detection. In Proceedings of the IEEE/CVF International Conference on Computer Vision, 13053--13064

    2021

  11. [11]

    Dai, S.; Qu, Y.; Shen, Y.; Zhang, S.; and Cao, L. 2026. PAR3D: A Unified 3D-MLLM with Part-Aware Representation for Scene Understanding. arXiv preprint arXiv:2606.06485

    Pith/arXiv arXiv 2026

  12. [12]

    Deng, J.; He, T.; Jiang, L.; Wang, T.; Dayoub, F.; and Reid, I. 2025. 3d-llava: Towards generalist 3d lmms with omni superpoint transformer. In Proceedings of the Computer Vision and Pattern Recognition Conference, 3772--3782

    2025

  13. [13]

    Fan, Z.; Zhang, J.; Li, R.; Zhang, J.; Chen, R.; Hu, H.; Wang, K.; Wang, P.; Qu, H.; Zhou, S.; et al. 2026. Vlm-3r: Vision-language models augmented with instruction-aligned 3d reconstruction. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 31054--31065

    2026

  14. [14]

    C.; Morerio, P.; and Murino, V

    Garcia, N. C.; Morerio, P.; and Murino, V. 2018. Modality distillation with multiple stream networks for action recognition. In Proceedings of the European Conference on Computer Vision (ECCV), 103--118

    2018

  15. [15]

    Hinton, G.; Vinyals, O.; and Dean, J. 2015. Distilling the knowledge in a neural network. arXiv preprint arXiv:1503.02531

    Pith/arXiv arXiv 2015

  16. [16]

    Hong, Y.; Zhen, H.; Chen, P.; Zheng, S.; Du, Y.; Chen, Z.; and Gan, C. 2023. 3d-llm: Injecting the 3d world into large language models. Advances in Neural Information Processing Systems, 36: 20482--20494

    2023

  17. [17]

    Huang, H.; Chen, Y.; Wang, Z.; Huang, R.; Xu, R.; Wang, T.; Liu, L.; Cheng, X.; Zhao, Y.; Pang, J.; et al. 2024. Chat-scene: Bridging 3d scene and large language models with object identifiers. Advances in Neural Information Processing Systems, 37: 113991--114017

    2024

  18. [18]

    Huang, X.; Wu, J.; Xie, Q.; and Han, K. 2026. 3drs: Mllms need 3d-aware representation supervision for scene understanding. Advances in Neural Information Processing Systems, 38: 67961--67988

    2026

  19. [19]

    Jain, A.; Gkanatsios, N.; Mediratta, I.; and Fragkiadaki, K. 2022. Bottom up top down detection transformers for language grounding in images and point clouds. In European Conference on Computer Vision, 417--433. Springer

    2022

  20. [20]

    Ju, T.; Sun, W.; Du, W.; Yuan, X.; Ren, Z.; and Liu, G. 2024. How large language models encode context knowledge? a layer-wise probing study. In Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024), 8235--8246

    2024

  21. [21]

    Li, J.; Li, D.; Savarese, S.; and Hoi, S. 2023. Blip-2: Bootstrapping language-image pre-training with frozen image encoders and large language models. In International conference on machine learning, 19730--19742. PMLR

    2023

  22. [22]

    Liu, H.; Li, C.; Wu, Q.; and Lee, Y. J. 2023. Visual instruction tuning. Advances in neural information processing systems, 36: 34892--34916

    2023

  23. [23]

    Liu, Z.; Qi, X.; and Fu, C.-W. 2021. 3d-to-2d distillation for indoor scene parsing. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 4464--4474

    2021

  24. [24]

    Lopez-Paz, D.; Bottou, L.; Sch \"o lkopf, B.; and Vapnik, V. 2015. Unifying distillation and privileged information. arXiv preprint arXiv:1511.03643

    Pith/arXiv arXiv 2015

  25. [25]

    C.; and Fei-Fei, L

    Luo, Z.; Hsieh, J.-T.; Jiang, L.; Niebles, J. C.; and Fei-Fei, L. 2018. Graph distillation for action detection with privileged modalities. In Proceedings of the European conference on Computer Vision (ECCV), 166--183

    2018

  26. [26]

    Ma, X.; Yong, S.; Zheng, Z.; Li, Q.; Liang, Y.; Zhu, S.-C.; and Huang, S. 2022. Sqa3d: Situated question answering in 3d scenes. arXiv preprint arXiv:2210.07474

    Pith/arXiv arXiv 2022

  27. [27]

    Vapnik, V.; and Vashist, A. 2009. A new learning paradigm: Learning using privileged information. Neural networks, 22(5-6): 544--557

    2009

  28. [28]

    Wu, X.; Liang, D.; Feng, T.; Xia, K.; Zhang, Y.; Li, X.; Tan, X.; and Bai, X. 2026. Generation Models Know Space: Unleashing Implicit 3D Priors for Scene Understanding. arXiv preprint arXiv:2603.19235

    Pith/arXiv arXiv 2026

  29. [29]

    F.; and Chai, J

    Yang, J.; Chen, X.; Madaan, N.; Iyengar, M.; Qian, S.; Fouhey, D. F.; and Chai, J. 2025. 3d-grand: A million-scale dataset for 3d-llms with better grounding and less hallucination. In Proceedings of the Computer Vision and Pattern Recognition Conference, 29501--29512

    2025

  30. [30]

    Yeh, C.-H.; Qian, S.; Wang, M.; Ma, Y.; Tighe, J.; and Xiao, F. 2026. Beyond 3D VQAs: Injecting 3D Spatial Priors into Vision-Language Models for Enhanced Geometric Reasoning. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 16723--16733

    2026

  31. [31]

    Zhang, J.; Zhou, S.; Liu, B.; Kadambi, A.; and Fan, Z. 2026. Spatialstack: Layered geometry-language fusion for 3d vlm spatial reasoning. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 38678--38688

    2026

  32. [32]

    Zhang, Y.; Wu, J.; Li, W.; Li, B.; Ma, Z.; Liu, Z.; and Li, C. 2024. Llava-video: Video instruction tuning with synthetic data. arXiv preprint arXiv:2410.02713

    Pith/arXiv arXiv 2024

  33. [33]

    Zheng, D.; Huang, S.; and Wang, L. 2025. Video-3d llm: Learning position-aware video representation for 3d scene understanding. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 8995--9006

    2025

  34. [34]

    Zhu, C.; Wang, T.; Zhang, W.; Pang, J.; and Liu, X. 2025. Llava-3d: A simple yet effective pathway to empowering lmms with 3d capabilities. In Proceedings of the IEEE/CVF International Conference on Computer Vision, 4295--4305

    2025

  35. [35]

    Zhu, Z.; Ma, X.; Chen, Y.; Deng, Z.; Huang, S.; and Li, Q. 2023. 3d-vista: Pre-trained transformer for 3d vision and text alignment. In Proceedings of the IEEE/CVF International Conference on Computer Vision, 2911--2921

    2023