Pith. sign in

REVIEW 4 major objections 6 minor 34 references

NEO edits one neural scene scan with language so a robot can predict many object moves without rescanning.

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 · grok-4.5

2026-07-31 12:10 UTC pith:XTARYF45

load-bearing objection Solid RA-L-style systems paper: permanent language-guided NeRF edits without rescanning, a real registered benchmark, and clear wins over the right baselines—scope is controlled tabletop, and the Out metrics partly ride easy surfaces. the 4 major comments →

arxiv 2607.24538 v1 pith:XTARYF45 submitted 2026-07-27 cs.RO

NEO: NeRF It Once, Edit It Many Times for Continuous Object Manipulation

Miko{\l}aj Zieli\'nski , David Hall , Dominik Belter , Peyman Moghadam This is my paper
classification cs.RO
keywords NeRF editingrobotic manipulationlanguage-guided scene editingknowledge distillationobject removalmultiview inpaintingNEO-Datasetcontinuous object manipulation
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.

Every robot grasp changes the scene, so neural radiance fields usually have to be rebuilt after each action. NEO claims a single initial scan is enough: given a language prompt, it removes the named object by resampling rays around its box, fills the newly exposed surfaces with multiview-consistent progressive inpainting, then reinserts the object at a planned pose by distilling one student network from the original and cleaned teachers. The new weights are permanent, so the same model accepts the next language command without a fresh scan. The authors also release NEO-Dataset—registered before-and-after tabletop captures—so edits can be scored against real geometry. On object-removal and pick-and-place tests the pipeline beats prior editors on image and depth metrics and supports multi-stage assembly from one scan.

Core claim

Language-guided removal by neural-field resampling, multiview progressive inpainting of revealed regions, and teacher–student weight distillation that composes the original and cleaned NeRFs produce persistent, view-consistent scene edits. Those edits match real post-manipulation geometry closely enough that a robot can plan successive grasps from the updated model alone, without rescanning.

What carries the argument

Region-wise teacher–student distillation: outside the object box the original NeRF supervises, inside the removal box the cleaned NeRF supervises, and at the relocated pose the original object is supervised under the planned motion—yielding one student whose weights permanently encode the reconfigured scene.

Load-bearing premise

That a diffusion inpainter run on a few virtual object-centered views, followed by mask-conditioned NeRF training and a density penalty inside the old box, reconstructs surfaces the cameras never saw well enough for the distilled model to match the real table after the robot moves the object—and that this stays true across several edits in a row.

What would settle it

Execute the full pick-and-place pipeline on the NEO-Dataset scenes and check whether depth RMSE and photometric reprojection error inside the Out and In masks remain at or below the reported levels and better than Seal-3D; systematic floaters, ghost objects, or wrong table texture after physical moves under the same capture protocol would falsify the claim.

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

If this is right

  • Robots can plan multi-step assemblies from one scan by predicting each intermediate scene state before acting.
  • Language fields remain queryable after every edit, so the same text interface works across a whole sequence.
  • Permanent weight updates remove the need to re-optimize or re-scan between actions.
  • NEO-Dataset supplies a shared, registered yardstick for scoring future NeRF editors on real manipulation outcomes.
  • Batch edits and small spatial offsets are supported inside the same pipeline.

Where Pith is reading between the lines

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

  • If inpainting quality holds under clutter or mild specular lighting, the same loop could act as a cheap predictive world model for model-based control without a physics simulator.
  • The resampling-plus-distillation pattern may transfer to other implicit scene models that currently degrade under repeated edits.
  • Likely failure modes concentrate on thin or transparent objects and on large viewpoint gaps that leave bigger unobserved regions than the virtual hemisphere covers.
  • Pairing NEO’s predicted frames with real post-action images could automatically label residual correction policies.

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

4 major / 6 minor

Summary. The paper presents NEO, a language-guided NeRF editing pipeline for robotic manipulation consisting of three stages: (1) object removal via ray-segment resampling that excludes the target bounding box from volumetric sampling, (2) multiview-consistent progressive diffusion inpainting on virtual object-centered views combined with mask-conditioned NeRF refinement and a density regularizer inside the removal box, and (3) teacher–student weight distillation that composes the original field (outside the box), the cleaned field (inside the box), and the transformed original field (relocated object) into a persistent edited model, with the language field subsequently updated from rendered views. The authors introduce NEO-Dataset, twelve registered pre/post-edit tabletop scenes captured with robot-mounted cameras, and report improvements over DFF, NeRFiller, and Seal-3D on five removal scenes (Table I), over Seal-3D on six pick-and-place scenes (Table II), an ablation (Table III), and a qualitative four-step assembly sequence (Fig. 10).

Significance. If the results hold, the paper makes two useful contributions to robot manipulation with neural fields: (i) a working, persistent weight-level NeRF editing pipeline that supports language-guided removal, relocation, and sequential assembly without re-scanning, demonstrated on a real Franka/UR5 setup; and (ii) NEO-Dataset, a registered pre/post-edit benchmark with real ground-truth post-manipulation scans — a genuinely scarce resource that enables quantitative evaluation of scene editing against physical reality rather than synthetic proxies. The code release (project page) and the use of external ground truth plus independent baselines (DFF, NeRFiller, Seal-3D) are real strengths. The impact is bounded, however, by the small evaluation scale and by the fact that the headline masked-region metrics are demonstrated only in a regime (uniform tabletop, controlled illumination) where the unobserved-surface reconstruction problem is nearly trivial.

major comments (4)
  1. [Tables II and III, §V.B–V.C] Table II vs. Table III, NEO full-model rows: for the same pick-and-place task, Table II reports NEO depth RMSE Out/In of 0.141/0.182 m, while Table III reports the full model at 0.056/0.033 m — a roughly 3-5x discrepancy. PSNR and SSIM rows agree (21.99/19.26, 0.783/0.619), so the tables appear to refer to the same evaluation. Either the RMSE figures come from different protocols/subsets (in which case this must be stated) or one row is misreported. This needs to be reconciled, since the RMSE numbers are cited as evidence of geometric consistency.
  2. [§III, §IV.A, Table I] The Out-region metrics (Table I) evaluate reconstruction of surfaces that were never observed in the original scan (acknowledged in §II.B). The reported Out PSNR of 25.43 dB is achieved on scenes with plain, low-texture tabletops under controlled illumination (§III), where the hidden surface is near-deterministic and a diffusion inpainter can match ground truth essentially by generating a uniform tabletop. The current benchmark therefore cannot distinguish faithful reconstruction of unobserved geometry from a lucky uniform hallucination, and it is unknown whether the margin over NeRFiller/DFF survives when the occluded surface carries information (printed pattern, textured mat, a partially hidden second object). This does not invalidate the relative comparisons — all baselines face the same conditions — but the paper should (a) state this limitation explicitly, and (b) ideally include at
  3. [§II.E, §V.D, Fig. 10] The title and central claim concern continuous, sequential editing ('Edit It Many Times'), yet the only sequential-edit evidence is the qualitative four-step assembly sequence in Fig. 10, with no quantitative metrics at intermediate stages. The reader cannot assess whether artifacts compound across edits — the single most important correctness-risk for the claimed use case (each edit's inpainted regions become the next edit's 'observed' content). Please add per-step quantitative evaluation (PSNR/SSIM/RMSE against the real post-action scans, which the authors appear to possess per Fig. 10's bottom row) for at least the assembly sequence, or soften the continuous-manipulation claim.
  4. [Tables I–II, §V] Evaluation statistics: five removal scenes (Table I) and six manipulation scenes (Table II) are reported as bare averages with no per-scene breakdown or variance. With n=5-6, 'consistently outperforms' (§V.A) is not established by the presented data; e.g., NEO's full-image removal PSNR margin over Seal-3D is 27.20 vs 27.08, well within plausible per-scene noise. Please report per-scene results (the dataset is small enough for a full table) or win/tie/loss counts per scene, at minimum for the claims where margins are small.
minor comments (6)
  1. [§II.B, Eqs. (4)-(5)] Notation in Eq. (4)-(5): Lreg is defined with (σ̃ᵢ − σ̃′ᵢ)² where σ̃′ = σ̃ outside b, so the loss vanishes outside the box by construction — fine, but the text says it 'applies only to samples within b', which is a consequence of the definition, not an additional property. Also σ̃ vs. σ̃′ is easy to misread; consider writing the loss explicitly as a sum over samples in b.
  2. [§II.C, Eq. (6)] Eq. (6): the piecewise definition of Ldistill does not state when each branch applies (the text says 'initial phase' vs. 'second phase' but the equation itself is unlabeled), and the variable n is reused both for pixel count here and for the re-inpainting interval in §II.B.
  3. [§V.A–V.B] Timing claims (6 min inpainting vs. 1.5 h NeRFiller; 1.5 vs. 3 min for manipulation) lack hardware specifications and details of baseline configurations. Please report GPU model and confirm baselines were run with recommended settings.
  4. [§II.D] §II.D: the claim that language-field retraining 'converges in a small number of iterations' is offered as a hypothesis; either quantify it (iterations/wall-clock, or downstream localization accuracy after several edits) or temper the statement.
  5. [§IV] Erep is cited to LSD-SLAM [34] but never defined; a one-line definition would help readers unfamiliar with the photometric reprojection error.
  6. [Throughout] Typos/formatting: 'we presentNEO' (p.1); inconsistent capitalization 'Nerfiller' vs. 'NeRFiller' (§IV and ref. [17]); Fig. 5's caption describes (b) as showing a hallucinated region but this is easy to misread as a method output rather than a motivating failure case — please clarify.

Circularity Check

0 steps flagged

No circularity: empirical NeRF-editing pipeline validated against external post-edit scans and independent baselines.

full rationale

NEO is a systems/methods paper whose load-bearing claims are empirical (outperformance on object removal and pick-and-place vs DFF, NeRFiller, Seal-3D; multi-stage assembly without re-scan). The pipeline—ray resampling outside a language-localized box, diffusion inpainting on virtual views with mask-conditioned NeRF refinement (Eqs. 3–5), and region-wise teacher–student distillation (Eq. 6)—optimizes standard reconstruction/distillation losses; none of these objectives redefine the reported metrics (PSNR/SSIM/depth RMSE/E_rep) by construction. Evaluation uses registered real post-manipulation images from NEO-Dataset as external ground truth, not self-generated targets. Citations to prior editing methods are baselines or related work, not uniqueness theorems or ansatzes that force the result. Concerns that unobserved ‘Out’ surfaces are near-uniform tabletops (so high PSNR is easy) are generalization/correctness issues, not circular derivation. No fitted input is relabeled a prediction; no step reduces Eq. X to Eq. Y by definition. Score 0 is the honest finding.

Axiom & Free-Parameter Ledger

4 free parameters · 5 axioms · 2 invented entities

Central claims rest on standard NeRF/volumetric rendering and language-field machinery, plus domain assumptions about tabletop capture and diffusion inpainting quality. Free parameters are loss weights and inpainting schedule knobs. No new physical entities; invented items are methodological modules and the benchmark.

free parameters (4)
  • λ_rec, λ_reg (masked training weights)
    Weighting constants in L_masked for reconstruction vs density regularization inside the removal box; chosen by authors, not derived.
  • inpainting strength schedule ξ = reduced each inpainting iteration (value unspecified)
    Progressively reduced during iterative virtual-view inpainting; controls how much of the current render is preserved vs regenerated.
  • virtual hemisphere view layout = four opposing azimuths + varying elevations
    Number/placement of object-centered virtual cameras (four opposing azimuths, varying elevations, short spiral) is a design choice affecting supervision density.
  • distillation stage sample/ray counts and loss switch
    Two-phase L_distill (per-sample density/color then RGB+depth image losses) and focus on virtual views are hand-designed training choices.
axioms (5)
  • domain assumption Standard NeRF volumetric rendering and hierarchical sampling correctly represent continuous scene density and appearance from multi-view images.
    Foundation of F_Θ and all edit stages (Sec. II, Mildenhall et al.).
  • domain assumption CLIP/DINO-style language features embedded in a language field localize the prompted object well enough to fit an oriented 3D bounding box for removal and grasp planning.
    Object selection via cosine similarity clustering (Sec. II intro); inherits LERF/DFF-style assumptions.
  • domain assumption A 2D diffusion inpainting model applied to 2×2 multiview composites yields completions that are sufficiently cross-view consistent to supervise a 3D NeRF of previously occluded surfaces.
    Core of Sec. II.B progressive inpainting; failure would break Out-region geometry metrics.
  • domain assumption Rigid object motion and controlled tabletop illumination (limited specular/lighting change) make pre/post registered scans a fair ground truth for edited NeRF quality.
    NEO-Dataset design (Sec. III); evaluation assumes appearance change is dominated by object pose, not lighting.
  • ad hoc to paper Region-wise composition of teacher fields (original outside box, cleaned inside removal box, transformed original for relocated object) defines a consistent target for student distillation.
    Sec. II.C multi-teacher supervision scheme specific to NEO's reconfiguration module.
invented entities (2)
  • NEO editable pipeline (resampling + progressive mask-conditioned inpainting + multi-teacher student NeRF) independent evidence
    purpose: Permanently update NeRF and language field weights to model post-manipulation scenes without rescanning.
    Methodological system composed from known components; evaluated empirically rather than postulated as a new physical object.
  • NEO-Dataset independent evidence
    purpose: Provide registered pre/post tabletop scans for quantitative NeRF editing evaluation in manipulation settings.
    New benchmark asset claimed as first of its kind for this use case; evidence is the described capture protocol and reported metrics.

pith-pipeline@v1.2.0-grok45-kimik3 · 16852 in / 3479 out tokens · 55180 ms · 2026-07-31T12:10:36.823489+00:00 · methodology

0 comments
read the original abstract

In this paper, we present NEO, a unified framework providing language-guided NeRF editing for robotic manipulation. Our paper introduces (i) a language-guided object removal that combines neural field resampling with multiview-consistent progressive inpainting, (ii) a direct NeRF weight editing method utilizing knowledge distillation, composing original and edited NeRFs via a teacher-student model, enabling coherent modeling of future scene states before a robot executes an action, and (iii) the first benchmark (NEO-Dataset) for quantitatively evaluating NeRF scene editing methods suitable for robot manipulation. We show that our approach outperforms state-of-the-art baselines in scene editing tasks, including object removal and pick-and-place robotic experiments, yielding visually coherent and geometrically consistent edits that reduce artifacts commonly introduced by prior methods.

Figures

Figures reproduced from arXiv: 2607.24538 by David Hall, Dominik Belter, Miko{\l}aj Zieli\'nski, Peyman Moghadam.

Figure 1
Figure 1. Figure 1: Example assembly task: given language prompts and a single initial [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Architecture of the proposed editable NEO model and its role within the robotic manipulation system. A set of camera images is used to build the [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Comparison of our neural field resampling method to the DFF [26] [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Proposed NeRF resampling method. The yellow color shows the [PITH_FULL_IMAGE:figures/full_fig_p003_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: (a) Images of the scene before and (b) after removing the selected [PITH_FULL_IMAGE:figures/full_fig_p004_5.png] view at source ↗
Figure 7
Figure 7. Figure 7: Object reconfiguration module for preservation of the structural [PITH_FULL_IMAGE:figures/full_fig_p005_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Example scenes after removing the selected object: (a) original scenes with object, (b) ground truth images after object removal, (c) DFF [26], (d) [PITH_FULL_IMAGE:figures/full_fig_p006_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Example images generated by (a) NEO and (b) Seal-3D [19] after [PITH_FULL_IMAGE:figures/full_fig_p007_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Assembly experiment showing NEO output (top row), which predicts the future state of the scene based on the initial scene (a) and a series of text [PITH_FULL_IMAGE:figures/full_fig_p008_10.png] 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

34 extracted references · 1 linked inside Pith

  1. [1]

    NeRF: Representing scenes as neural radiance fields for view synthesis

    B. Mildenhall, P. P. Srinivasanet al., “NeRF: Representing scenes as neural radiance fields for view synthesis.” inECCV, vol. 12346, 2020, pp. 405–421

    2020

  2. [2]

    NeRFs in robotics: A survey,

    G. Wang, L. Panet al., “NeRFs in robotics: A survey,”IJRR, 2024

    2024

  3. [3]

    Neural fields meet explicit geometric repre- sentations for inverse rendering of urban scenes

    Z. Wang, T. Shenet al., “Neural fields meet explicit geometric repre- sentations for inverse rendering of urban scenes.” inCVPR, 2023, pp. 8370–8380

    2023

  4. [4]

    Silvr: Scalable lidar-visual reconstruction with neural radiance fields for robotic inspection,

    Y . Tao, Y . Bhalgatet al., “Silvr: Scalable lidar-visual reconstruction with neural radiance fields for robotic inspection,” inICRA, 2024, pp. 17 983–17 989

    2024

  5. [5]

    NeRF2Real: Sim2real transfer of vision- guided bipedal motion skills using neural radiance fields

    A. Byravan, J. Humpliket al., “NeRF2Real: Sim2real transfer of vision- guided bipedal motion skills using neural radiance fields.” inICRA, 2023, pp. 9362–9369

    2023

  6. [6]

    Implicit grasp diffusion: Bridging the gap between dense prediction and sampling-based grasping,

    P. Song, P. Li, and R. Detry, “Implicit grasp diffusion: Bridging the gap between dense prediction and sampling-based grasping,” inCoRL, 2024

    2024

  7. [7]

    Language embedded radiance fields for zero-shot task-oriented grasping,

    A. Rashid, S. Sharmaet al., “Language embedded radiance fields for zero-shot task-oriented grasping,” inCoRL, 2023

    2023

  8. [8]

    OpenNeRF: Open set 3D neural scene segmentation with pixel-wise features and rendered novel views

    F. Engelmann, F. Manhardtet al., “OpenNeRF: Open set 3D neural scene segmentation with pixel-wise features and rendered novel views.” inICLR, 2024

    2024

  9. [9]

    CLA-NeRF: Category-level articulated neural radiance field

    W.-C. Tseng, H.-J. Liaoet al., “CLA-NeRF: Category-level articulated neural radiance field.” inICRA, 2022, pp. 8454–8460

    2022

  10. [10]

    Dex-NeRF: Using a neural radiance field to grasp transparent objects

    J. Ichnowski, Y . Avigalet al., “Dex-NeRF: Using a neural radiance field to grasp transparent objects.” inCoRL, vol. 164, 2021, pp. 526–536

    2021

  11. [11]

    NeRF is a valuable assistant for 3D gaussian splatting,

    S. Fang, I.-C. Shenet al., “NeRF is a valuable assistant for 3D gaussian splatting,” inICCV, 2025, pp. 26 230–26 240

    2025

  12. [12]

    Radiance fields for robotic teleoperation,

    M. Wilder-Smith, V . Patil, and M. Hutter, “Radiance fields for robotic teleoperation,” inIROS, 2024, pp. 13 861–13 868

    2024

  13. [13]

    Occlusion-free scene recovery via neural radiance fields

    C. Zhu, R. Wanet al., “Occlusion-free scene recovery via neural radiance fields.” inCVPR. IEEE, 2023, pp. 20 722–20 731

    2023

  14. [14]

    LERF: Language embedded radiance fields

    J. Kerr, C. M. Kimet al., “LERF: Language embedded radiance fields.” inICCV, 2023, pp. 19 672–19 682

    2023

  15. [15]

    Distilled feature fields enable few-shot language-guided manipulation,

    W. Shen, G. Yanget al., “Distilled feature fields enable few-shot language-guided manipulation,” inCoRL, 2023

    2023

  16. [16]

    Fast sparse view guided NeRF update for object reconfigurations

    Z. Lu, J. Yeet al., “Fast sparse view guided NeRF update for object reconfigurations.”CoRR, vol. abs/2403.11024, 2024

    Pith/arXiv arXiv 2024

  17. [17]

    Nerfiller: Completing scenes via gener- ative 3d inpainting,

    E. Weber, A. Holynskiet al., “Nerfiller: Completing scenes via gener- ative 3d inpainting,” inCVPR, 2024, pp. 20 731–20 741

    2024

  18. [18]

    ReplaceAnything3D: Text-guided object replacement in 3D scenes with compositional scene representa- tions

    E. Bartrum, T. Nguyen-Phuocet al., “ReplaceAnything3D: Text-guided object replacement in 3D scenes with compositional scene representa- tions.” inNeurIPS, 2024

    2024

  19. [19]

    Seal-3D: Interactive pixel-level editing for neural radiance fields,

    X. Wang, J. Zhuet al., “Seal-3D: Interactive pixel-level editing for neural radiance fields,” inICCV, 2023, pp. 17 637–17 647

    2023

  20. [20]

    NeuralEditor: Editing neural radiance fields via manipulating point clouds

    J.-K. Chen, J. Lyu, and Y .-X. Wang, “NeuralEditor: Editing neural radiance fields via manipulating point clouds.” inCVPR, 2023, pp. 12 439–12 448

    2023

  21. [21]

    NeRF-Editing: Geometry editing of neural radiance fields

    Y .-J. Yuan, Y .-T. Sunet al., “NeRF-Editing: Geometry editing of neural radiance fields.” inCVPR, 2022, pp. 18 332–18 343

    2022

  22. [22]

    Learning unified decompositional and compositional NeRF for editable novel view synthesis

    Y . Wang, W. Wu, and D. Xu, “Learning unified decompositional and compositional NeRF for editable novel view synthesis.” inICCV. IEEE, 2023, pp. 18 201–18 210

    2023

  23. [23]

    Reg-NF: Efficient registration of implicit surfaces within neural fields,

    S. Hausler, D. Hallet al., “Reg-NF: Efficient registration of implicit surfaces within neural fields,” inICRA, 2024, pp. 15 409–15 415

    2024

  24. [24]

    Removing objects from neural radiance fields,

    S. Weder, G. Garcia-Hernandoet al., “Removing objects from neural radiance fields,” inCVPR, 2023

    2023

  25. [25]

    High-resolution image synthesis with latent diffusion models,

    R. Rombach, A. Blattmannet al., “High-resolution image synthesis with latent diffusion models,” inCVPR, 2022, pp. 10 684–10 695

    2022

  26. [26]

    Decomposing NeRF for editing via feature field distillation

    S. Kobayashi, E. Matsumoto, and V . Sitzmann, “Decomposing NeRF for editing via feature field distillation.” inNeurIPS, 2022

    2022

  27. [27]

    DATENeRF: Depth-aware text-based editing of NeRFs,

    S. Rojas, J. Philipet al., “DATENeRF: Depth-aware text-based editing of NeRFs,” inECCV, 2024, pp. 267–284

    2024

  28. [28]

    ViCA-NeRF: view-consistency-aware 3D editing of neural radiance fields,

    J. Dong and Y .-X. Wang, “ViCA-NeRF: view-consistency-aware 3D editing of neural radiance fields,” inNeurIPS, 2023

    2023

  29. [29]

    Physically embodied gaussian splatting: A realtime correctable world model for robotics,

    J. Abou-Chakra, K. Ranaet al., “Physically embodied gaussian splatting: A realtime correctable world model for robotics,” inCoRL, 2024

    2024

  30. [30]

    Splat-mover: Multi-stage, open- vocabulary robotic manipulation via editable gaussian splatting,

    O. Shorinwa, J. Tuckeret al., “Splat-mover: Multi-stage, open- vocabulary robotic manipulation via editable gaussian splatting,” in CoRL, 2024

    2024

  31. [31]

    Is there anything left? measuring semantic residuals of objects removed from 3D gaussian splatting,

    S. Kocour, A. Benbihiet al., “Is there anything left? measuring semantic residuals of objects removed from 3D gaussian splatting,” 2025

    2025

  32. [32]

    AnyGrasp: Robust and efficient grasp perception in spatial and temporal domains,

    H.-S. Fang, C. Wanget al., “AnyGrasp: Robust and efficient grasp perception in spatial and temporal domains,”T-RO, 2023

    2023

  33. [33]

    Benchmarking in manipulation research: Using the Yale-CMU-Berkeley object and model set

    B. C ¸ alli, A. Walsmanet al., “Benchmarking in manipulation research: Using the Yale-CMU-Berkeley object and model set.”IEEE Robotics Autom. Mag., vol. 22, no. 3, pp. 36–52, 2015

    2015

  34. [34]

    LSD-SLAM: Large-scale direct monocular SLAM,

    J. Engel, T. Sch ¨ops, and D. Cremers, “LSD-SLAM: Large-scale direct monocular SLAM,” inECCV, 2014, pp. 834–849

    2014