Pith. sign in

REVIEW 4 major objections 9 minor 22 references

Adding web images to text queries improves long-tail object finding in mapped environments without retraining the navigator.

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 ↗

Web-image-enriched queries improve open-vocabulary map grounding and ObjectNav success, especially for fine-grained categories, with gains limited by downstream detectors.

T0 review reviewed 2026-07-31 challenge →

load-bearing objection Solid plug-and-play engineering plus a real long-tail benchmark; headline grounding averages are inflated by a broken SigLIP text baseline, but the direction of the result and the detection-bottleneck diagnosis still hold. the 4 major comments →

arxiv 2607.25106 v1 pith:W47CHPYK submitted 2026-07-27 cs.CV

IMPRINT: Image-Conditioned Query Enrichment for Long-Tail Object Goal Navigation

Jelin Raphael Akkara , Filippo Ziliotto , Luciano Serafini , Lamberto Ballan , Tommaso Campari This is my paper
classification cs.CV
keywords Object Goal Navigationqueryable semantic mapsvision-language modelsimage-conditioned querieslong-tail recognitionzero-shot navigationHSSD-rareweb image retrieval
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 reading

Queryable semantic maps let robots find objects from open-vocabulary text, but a category name alone often fails when the target is a fine-grained or product-level subcategory. IMPRINT enriches that text query with a handful of web-retrieved images, encodes them with the same vision-language model used for the map, and averages the resulting similarity maps to pick a better place to go. The method is zero-shot and plug-and-play: it does not train anything or change the navigation policy. On a synonym-heavy split and on a new long-tail benchmark built from richly annotated synthetic scenes, image-conditioned queries raise isolated grounding success and produce end-to-end navigation gains. The paper further shows that those localization gains only fully become navigation success when the downstream object detector is also strong, identifying detection quality as a systems bottleneck for long-tail embodied search.

Core claim

Enriching a text object query with web-sourced image embeddings, matching each against a queryable semantic map, and aggregating the similarity maps consistently improves zero-shot object grounding and yields navigation gains on both synonym-level and true subcategory long-tail ObjectNav, without any training or policy change; converting those gains fully into success still depends on downstream detection quality.

What carries the argument

IMPRINT: retrieve N web images for the query (with head-noun metadata gating), encode them, produce one cosine-similarity map per image over the stored feature map, and average those maps (optionally with the text map) to obtain the target localization used by an unchanged navigation stack.

Load-bearing premise

Web search plus simple metadata filtering must return visual examples that are relevant enough that averaging their embeddings against the map is a trustworthy location signal rather than noise.

What would settle it

On HSSD-rare-style subcategory episodes, replace retrieved images with random or off-topic images (or disable gating) and check whether success rate and distance-to-goal still beat text-only; if they do not fall, the claimed benefit of image conditioning is not doing the work.

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

If this is right

  • Text-only queryable maps underperform as soon as targets move from broad classes to fine-grained subcategories; image exemplars close much of that gap.
  • The same plug-in can sit on map-based pipelines (frontier similarity plus PointGoal) and on direct policy conditioning without redesigning the controller.
  • Isolated grounding gains do not automatically equal navigation gains; detector quality becomes the binding constraint under long-tail conditions.
  • Conditioning the open-vocabulary detector on the same retrieved images unlocks further end-to-end success on both OVON-syn and HSSD-rare.
  • HSSD-rare supplies a reusable testbed of co-occurring subcategory targets for measuring true long-tail ObjectNav rather than synonym rewording alone.

Where Pith is reading between the lines

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

  • Joint image conditioning of both the map query and the detector may be the practical default for long-tail embodied search, not an optional ablation.
  • Retrieval quality and domain shift of web images could become a first-class failure mode when the same method moves from synthetic scenes to messy real homes.
  • Patch-level encoders help on common categories but may need different aggregation or depth filtering when targets are rare product variants.
  • Viewpoint-generation pipelines like the one used for HSSD-rare could be reused to stress-test other open-vocabulary navigation stacks on the same subcategory ontology.
Share X Bluesky LinkedIn Reddit HN

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 / 9 minor

Summary. The paper introduces IMPRINT, a training-free, plug-and-play module for open-vocabulary ObjectNav built on queryable semantic maps: given a text goal, it retrieves N web images (gated by a head-noun metadata check), encodes them with the same VLM used for mapping, computes per-image cosine-similarity maps over the map, and averages them for goal localization. It is evaluated in a static grounding phase (BLIP2, SigLIP, SED encoders; OVON-syn and the new HSSD-rare benchmark) and an online navigation phase (ZSON, VLFM, OneMap), with ablations over aggregation rule, image count, and OVON splits; a failure analysis; a controlled detector intervention (IMPRINT*); and a 5-episode LoCoBot demo. The authors claim consistent grounding and navigation gains, larger under fine-grained specificity, and identify object detection as the key bottleneck in the long-tail regime. HSSD-rare contributes 1000 episodes over 20 parent categories / 559 subcategories in 17 scenes, with a documented viewpoint-generation pipeline.

Significance. If the numbers hold after baseline cleanup, IMPRINT is a useful, low-overhead addition to open-vocabulary ObjectNav: training-free, no policy modification, demonstrated across three encoders and three navigation stacks, with gains consistent in direction on two benchmarks. Two elements have value beyond the method itself: HSSD-rare, a genuinely new subcategory-level ObjectNav benchmark (1000 episodes, 559 fine-grained subcategories, documented viewpoint generation, OOV and LLM-scored rarity analyses), and the controlled detector intervention (IMPRINT*) that cleanly separates grounding gains from detection bottlenecks — an informative systems result. Regime-dependence is reported honestly (OVON-seen +1.01 SR). The method itself (retrieve–encode–average) is simple, so the benchmark and the grounding-vs-detection decomposition are the lasting contributions.

major comments (4)
  1. [§VI-C, Table I] The pooled gains quoted in the text (+10.54 SR on OVON-syn, +6.37 on HSSD-rare) are dominated by the SigLIP text-only rows (7.74 and 2.15 SR), which trail BLIP2 text-only (38.71/17.11) on identical maps and protocol by >30 points. SigLIP-style encoders are normally strong zero-shot classifiers when standard prompt ensembling is used, so this pattern suggests an under-configured text baseline (e.g., raw category strings without prompt templates) rather than an intrinsic deficit of text queries. Excluding SigLIP, per-encoder gains are +3.55 (BLIP2) and +7.42 (SED) on OVON-syn, and +2.58/+2.18 on HSSD-rare — positive but much smaller than the headline. Please disclose the per-encoder text-prompting protocol, re-report SigLIP with standard prompt ensembling, and present per-encoder deltas in the text rather than pooled averages that hinge on one anomalous row.
  2. [§VI-A/C, Tables I–III] No evaluation reports episode counts or statistical uncertainty. The OVON-syn static evaluation is described only as 'balanced across categories' (§VI-A) without n, and several online deltas are small enough to be sampling noise at typical ObjectNav episode counts (e.g., ZSON +0.47 SR and OneMap +0.47 SR on HSSD-rare, Table II). Please state n for every table (static and online, including the subsampling protocol for OVON splits) and report paired confidence intervals (e.g., bootstrap over episodes) or per-category success rates, at least for Tables I–III. The 'consistent improvement' claim currently rests partly on deltas within plausible noise.
  3. [§VI-C (Table III), §IV-D (Table II)] Two online configurations are not reproducible from the text. (a) IMPRINT* 'conditions YOLOWorld with the top-3 retrieved reference images,' but YOLOWorld is a text-prompted detector and §IV describes only map-query enrichment; the image-conditioning mechanism, detection thresholds, and top-3 selection rule are never specified, yet Table III carries the 'detection bottleneck' contribution. (b) ZSON builds no semantic map, so the similarity-map aggregation of §IV-B does not apply; please describe how image embeddings condition the ZSON policy (e.g., an averaged goal embedding) and whether N and the gating rule match the map-based setting.
  4. [§IV-A, Fig. 5] The method's premise — that metadata-gated web retrieval yields sufficiently relevant exemplars — is never measured directly. The n-ablation (saturation then decline) is consistent with retrieval noise, and the gating step itself is not ablated. A small relevance audit (e.g., human or VLM relevance labels on retrieved sets for a sample of categories) plus a gated-vs-ungated comparison would substantiate the premise. Relatedly, please release the cached retrieved image sets: search-engine drift otherwise makes the headline numbers difficult for others to reproduce.
minor comments (9)
  1. [Fig. 5] The caption states the encoder is 'fixed as BLIP2,' but the legend shows BLIP2, SigLIP, and SED; please reconcile. Also state whether n>15 was tested — the HSSD-rare optimum sits at the edge of the swept range (n=15).
  2. [Table V] The harmonic mean is computed over cosine similarities, which can be negative; the harmonic mean is well-defined only for positive values. Please clarify any shift/clamping — otherwise the harmonic/hybrid collapse (e.g., SigLIP HSSD-rare SR 1.23) may be a numerical artifact rather than evidence about aggregation quality.
  3. [Abstract / §I] 'github.io/IMPRINT' is not a resolvable repository URL. Please provide the actual link and specify the release scope (HSSD-rare episodes, generated viewpoints, cached retrievals, evaluation configs) and license.
  4. [Appendix C] With five episodes, 1/5 vs 4/5 is not a meaningful quantitative comparison; please frame the LoCoBot study strictly as a qualitative proof-of-concept and, if possible, include per-episode details (goal, outcome, failure cause).
  5. [Table III] On OVON-syn, IMPRINT* raises SR from 13.00 to 18.17 but worsens DTG (4.53 → 4.98). A one-sentence explanation (e.g., additional successes on farther goals) would preempt confusion.
  6. [§VI-A / §IV-B] SR is defined via the stop action (§III), which never occurs in the static phase; please define static-phase SR explicitly (top-ranked map cell within 1 m of the target) and state the map grid resolution, since 1 m success depends on cell size.
  7. [§VI-A/C] Terminology: 'OVON-syn,' 'OVON-synon.' (Table IV), and 'val seen synonyms' are used interchangeably; please unify. IMPRINT* is first used in Table III without a defining sentence in the text.
  8. [§VI-C] Category-wise gains (treadmill, clothing, bottle, table lamp) are described but not tabulated; a per-category table in the appendix would make the 'broadly distributed' claim checkable.
  9. [Appendix A / throughout] Formatting artifacts: '2,m', '1,m', '0.5,m' in Appendix A and several broken inline math expressions (e.g., observation dimensions in §IV-B) should be repaired in the final PDF.

Circularity Check

0 steps flagged

No circularity: empirical plug-and-play method evaluated on external geometric success, not a derivation that reduces to its inputs.

full rationale

IMPRINT is a zero-shot engineering contribution (web-image retrieval + cosine similarity-map averaging on frozen VLM features) plugged into existing ObjectNav pipelines. Success is defined externally by Habitat geometry (stop within 1 m of a target instance), not by re-using a fitted objective or a self-defined quantity. Aggregation (arithmetic mean of image-conditioned cosine maps) does not tautologically force higher SR; text-only and image-conditioned modes are compared on the same cached maps and held-out episodes. Ablations (n images, aggregation, detector conditioning) vary free design choices rather than renaming fits as predictions. Baselines (VLFM, OneMap, ZSON, BLIP2/SigLIP/SED) and the OVON split are external; HSSD-rare is a new benchmark constructed from HSSD assets with an independent viewpoint pipeline. There is no self-definitional loop, no fitted-parameter-as-prediction structure, no load-bearing uniqueness theorem from overlapping authors, and no ansatz smuggled in via self-citation. Skeptic concerns about SigLIP text-baseline strength and small residual deltas are validity/power issues, not circularity.

Axiom & Free-Parameter Ledger

4 free parameters · 4 axioms · 2 invented entities

Empirical systems paper. Load-bearing background is standard VLM cosine grounding and existing ObjectNav map pipelines; free choices are retrieval count, aggregation, depth masking, and gating heuristics. No new physical entities; HSSD-rare is a constructed benchmark, not a postulated mechanism.

free parameters (4)
  • number of retrieved images N = 10 (OVON-syn), 15 (HSSD-rare)
    Chosen by ablation; peaks at N=10 (OVON-syn) and N=15 (HSSD-rare). Central gains depend on operating near this range.
  • depth mask / max range for mapping = 4 m
    RGB depth-masked to 4 m and depth sensor capped at 4 m in static mapping to reduce projection noise; affects map quality and thus grounding scores.
  • similarity-map aggregation rule = arithmetic mean
    Arithmetic mean selected after comparing mean, harmonic mean, and hybrid; performance is sensitive to this choice (Table V).
  • metadata gating rule for web images = head-noun in ALT or URL
    Keep image only if head noun appears in ALT text or URL; ad hoc relevance filter that shapes the exemplar set.
axioms (4)
  • domain assumption Cosine similarity between VLM embeddings of queries (text or web images) and map cell features is a valid ranking signal for object location.
    Inherited from VLMaps/VLFM-style pipelines; entire static and online evaluation rests on it (Sec. IV-B).
  • domain assumption Web image search returns exemplars whose visual features transfer to the agent’s egocentric observations of the same category name.
    Stated motivation in Sec. I and IV-A; without domain transfer, image conditioning cannot help.
  • domain assumption Success within 1 m geodesic of any instance of the queried subcategory is the correct ObjectNav success criterion.
    Standard Habitat ObjectNav definition adopted in Sec. III; defines all SR/SPL numbers.
  • ad hoc to paper Prepending the head category to long-tail queries improves retrieval quality without changing the navigation task definition.
    Sec. IV-A retrieval detail specific to this method.
invented entities (2)
  • HSSD-rare benchmark independent evidence
    purpose: Provide ObjectNav episodes over fine-grained subcategories with generated viewpoints to stress long-tail grounding.
    Constructed dataset (1000 episodes, 559 subcategories, 17 scenes), not a physical entity; independent use is possible once released.
  • IMPRINT image-conditioned query module no independent evidence
    purpose: Aggregate web-image similarity maps (optionally with text) as a drop-in replacement for text-only map queries.
    Method artifact defined by the pipeline in Sec. IV; evidence is the reported experiments, not an external phenomenon.

reviewed 2026-07-31 · how reviews work

0 comments
Cite this review

Pith. "Pith review of IMPRINT: Image-Conditioned Query Enrichment for Long-Tail Object Goal Navigation." pith.science (2026). https://pith.science/paper/W47CHPYK

@misc{pith2026260725106,
  author       = {Pith},
  title        = {Pith review of: IMPRINT: Image-Conditioned Query Enrichment for Long-Tail Object Goal Navigation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/W47CHPYK}},
  note         = {Machine review of arXiv:2607.25106}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Embodied AI increasingly relies on queryable semantic maps built from pre-trained vision-language models to enable zero-shot Object Goal Navigation (ObjectNav). However, existing approaches typically depend on text-only queries, which become less reliable as semantic specificity increases toward fine-grained object categories. We introduce IMPRINT, a zero-shot plug-and-play framework that enriches textual object queries with web-sourced images to improve grounding in queryable maps. Retrieved images are encoded using a vision-language model, matched against the semantic map to produce similarity maps, and aggregated to yield context-aware localization. Notably, this requires no training or modification of the underlying navigation policy. To explicitly evaluate long-tail behavior, we present HSSD-rare, a new ObjectNav benchmark built on Habitat Synthetic Scenes and featuring semantically specific subcategories. Across both OVON and HSSD-rare, image-conditioned queries consistently improve object grounding and yield end-to-end navigation gains. Further analysis reveals that translating localization gains to navigation performance depends critically on downstream detection quality, highlighting a key systems bottleneck in long-tail embodied navigation.

Figures

Figures reproduced from arXiv: 2607.25106 by Filippo Ziliotto, Jelin Raphael Akkara, Lamberto Ballan, Luciano Serafini, Tommaso Campari.

Figure 1
Figure 1. Figure 1: We present IMPRINT, a simple yet effective plug-and-play method [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: IMPRINT pipeline. (left) RGB observations are encoded by a vision–language model and projected into a queryable semantic feature map. (right) Given a text query, relevant web images are retrieved and encoded; each embedding produces a similarity map over the stored feature map, and these maps are aggregated to yield a context-aware target localization. In the static phase, mapping is completed first and ob… view at source ↗
Figure 3
Figure 3. Figure 3: HSSD-rare semantic categories: Categories are organized hierarchically: each row corresponds to a main category, while columns illustrate long-tail subcategory variation. indoor–outdoor regions, yet the dataset does not provide explicit floor assignments or connectivity annotations. This ambiguity can disrupt frontier-based exploration, as agents may inadvertently transition between floors or encounter ina… view at source ↗
Figure 4
Figure 4. Figure 4: Viewpoint generation for HSSD-rare. For each target instance (red), valid viewpoints (yellow) are identified by enforcing visibility and distance constraints. Invalid examples (a,b) are rejected due to occlusion or excessive distance. Valid viewpoints (c,d) demonstrate spatial diversity across cluttered and open regions, preventing trivial start–goal configura￾tions. Episode Sampling. For each main categor… view at source ↗
Figure 5
Figure 5. Figure 5: Ablation with Number of Images: Object grounding success rate is measured as the number of images (k) is varied, keeping the feature encoder fixed as BLIP2. Optimal values are marked as vertical lines. detection. Moreover, evaluation on HSSD-rare may not fully reflect real-world long-tailed ObjectGoal Navigation, moti￾vating validation in more realistic settings. VII. CONCLUSION We introduced IMPRINT, a ze… view at source ↗
Figure 7
Figure 7. Figure 7: Examples of viewpoint validation using RGB and depth obser [PITH_FULL_IMAGE:figures/full_fig_p009_7.png] view at source ↗
Figure 6
Figure 6. Figure 6: Viewpoint-generation pipeline for a target object. (a) The target object’s location and dimensions are obtained, with its location shown in red. (b) A circle of radius 2 m is centered at the target, and points sampled at 0.5 m intervals along transitions from non-traversable to traversable space are retained as accessible boundary points, shown in blue. (c) Candidate viewpoints, shown in yellow, are genera… view at source ↗
Figure 8
Figure 8. Figure 8: LLM-based rarity distribution (1–5 scale) for HSSD-rare vs. OVON [PITH_FULL_IMAGE:figures/full_fig_p010_8.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

22 extracted references · 2 linked inside Pith

  1. [1]

    Learning transferable visual models from natural language su- pervision,

    A. Radford, J. W. Kim, C. Hallacy, A. Ramesh, G. Goh, S. Agarwal, et al., “Learning transferable visual models from natural language su- pervision,” inInternational Conference on Machine Learning (ICML), 2021

    2021

  2. [2]

    Blip-2: Bootstrapping language- image pre-training with frozen image encoders and large language models,

    J. Li, D. Li, S. Savarese, and S. Hoi, “Blip-2: Bootstrapping language- image pre-training with frozen image encoders and large language models,” inInternational conference on machine learning. PMLR, 2023

    2023

  3. [3]

    Sed: A simple encoder-decoder for open-vocabulary semantic segmentation,

    B. Xie, J. Cao, J. Xie, F. S. Khan, and Y . Pang, “Sed: A simple encoder-decoder for open-vocabulary semantic segmentation,” inPro- ceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2024

    2024

  4. [4]

    Vlfm: Vision- language frontier maps for zero-shot semantic navigation,

    N. Yokoyama, S. Ha, D. Batra, J. Wang, and B. Bucher, “Vlfm: Vision- language frontier maps for zero-shot semantic navigation,” in2024 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2024

    2024

  5. [5]

    Onemap to find them all: A spatial semantic map for multi-query object search,

    J. Buschet al., “Onemap to find them all: A spatial semantic map for multi-query object search,” inIEEE International Conference on Robotics and Automation (ICRA), 2024

    2024

  6. [6]

    Search and detect: Training-free long tail object detection via web- image retrieval,

    M. Sidhu, H. Chopra, A. Blume, J. Kim, R. G. Reddy, and H. Ji, “Search and detect: Training-free long tail object detection via web- image retrieval,” inProceedings of the Computer Vision and Pattern Recognition Conference, 2025, pp. 15 129–15 138

    2025

  7. [7]

    Visualwe- binstruct: Scaling up multimodal instruction data through web search,

    Y . Jia, J. Li, X. Yue, B. Li, P. Nie, K. Zou, and W. Chen, “Visualwe- binstruct: Scaling up multimodal instruction data through web search,” arXiv preprint arXiv:2503.10582, 2025

    Pith/arXiv arXiv 2025

  8. [8]

    Hm3d- ovon: A dataset and benchmark for open-vocabulary object goal navigation,

    N. Yokoyama, R. Ramrakhya, A. Das, D. Batra, and S. Ha, “Hm3d- ovon: A dataset and benchmark for open-vocabulary object goal navigation,” in2024 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). IEEE, 2024

    2024

  9. [9]

    Habitat synthetic scenes dataset (hssd-200): An analysis of 3d scene scale and realism tradeoffs for objectgoal navigation,

    M. Khanna, Y . Mao, H. Jiang, S. Haresh, B. Shacklett, D. Batra, A. Clegg, E. Undersander, A. X. Chang, and M. Savva, “Habitat synthetic scenes dataset (hssd-200): An analysis of 3d scene scale and realism tradeoffs for objectgoal navigation,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024

    2024

  10. [10]

    Matterport3d: Learning from rgb- d data in indoor environments,

    A. Chang, A. Dai, T. Funkhouser, M. Halber, M. Niebner, M. Savva, S. Song, A. Zeng, and Y . Zhang, “Matterport3d: Learning from rgb- d data in indoor environments,” inInternational Conference on 3D Vision (3DV), 2017

    2017

  11. [11]

    Gibson env: Real-world perception for embodied agents,

    F. Xia, A. R. Zamir, Z. He, A. Sax, J. Malik, and S. Savarese, “Gibson env: Real-world perception for embodied agents,” inProceedings of the IEEE conference on computer vision and pattern recognition, 2018

    2018

  12. [12]

    Vlmaps: Vision- language maps for robot navigation,

    C. Huang, O. Mees, A. Zeng, and W. Burgard, “Vlmaps: Vision- language maps for robot navigation,” inIEEE International Confer- ence on Robotics and Automation (ICRA), 2023

    2023

  13. [13]

    Openfusion: Real-time open-vocabulary 3d mapping and queryable scene representation,

    K. Yamazaki, T. Hanyu, K. V o, T. Pham, M. Tran, G. Doretto, A. Nguyen, and N. Le, “Openfusion: Real-time open-vocabulary 3d mapping and queryable scene representation,” inProceedings of ICRA, 2024

    2024

  14. [14]

    Sigmoid loss for language image pre-training,

    X. Zhai, B. Mustafa, A. Kolesnikov, and L. Beyer, “Sigmoid loss for language image pre-training,” inProceedings of the IEEE/CVF international conference on computer vision, 2023, pp. 11 975–11 986

    2023

  15. [15]

    Lseg: Language-driven semantic segmentation,

    B. Li, K. Q. Weinberger, S. Belongie, V . Koltun, and R. Ranftl, “Lseg: Language-driven semantic segmentation,” inInternational Conference on Learning Representations (ICLR), 2022

    2022

  16. [16]

    Openscene: 3d scene understanding with open vo- cabularies,

    S. Peng, K. Genova, C. Jiang, A. Tagliasacchi, M. Pollefeys, and T. Funkhouser, “Openscene: 3d scene understanding with open vo- cabularies,” inIEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2023

    2023

  17. [17]

    Scaling open-vocabulary object detection,

    M. Minderer, A. Gritsenko, and N. Houlsby, “Scaling open-vocabulary object detection,”Advances in Neural Information Processing Systems, vol. 36, 2023

    2023

  18. [18]

    Grounding dino: Marrying dino with grounded pre-training for open-set object detection,

    S. Liu, Z. Zeng, T. Ren, F. Li, H. Zhang, J. Yang, Q. Jiang, C. Li, J. Yang, H. Su,et al., “Grounding dino: Marrying dino with grounded pre-training for open-set object detection,” inEuropean Conference on Computer Vision. Springer, 2024

    2024

  19. [19]

    Yolo- world: Real-time open-vocabulary object detection,

    T. Cheng, L. Song, Y . Ge, W. Liu, X. Wang, and Y . Shan, “Yolo- world: Real-time open-vocabulary object detection,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2024, pp. 16 901–16 911

    2024

  20. [20]

    Deep long- tailed learning: A survey,

    Y . Zhang, B. Kang, B. Hooi, S. Yan, and J. Feng, “Deep long- tailed learning: A survey,”IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 45, no. 9, pp. 10 795–10 816, 2023

    2023

  21. [21]

    Habitat-matterport 3d dataset (hm3d): 1000 large-scale 3d environments for embodied ai,

    S. K. Ramakrishnan, A. Gokaslan, E. Wijmans, O. Maksymets, A. Clegg, J. Turner, E. Undersander, W. Galuba, A. Westbury, A. X. Chang,et al., “Habitat-matterport 3d dataset (hm3d): 1000 large-scale 3d environments for embodied ai,”arXiv preprint arXiv:2109.08238, 2021

    Pith/arXiv arXiv 2021

  22. [22]

    Zson: Zero-shot object-goal navigation using multimodal goal embed- dings,

    A. Majumdar, G. Aggarwal, B. Devnani, J. Hoffman, and D. Batra, “Zson: Zero-shot object-goal navigation using multimodal goal embed- dings,” inNeural Information Processing Systems (NeurIPS), 2022. APPENDIX A. HSSD-rare viewpoint generation pipeline To access the long-tail object categories in the HSSD dataset, we introduce a dedicated viewpoint-generatio...

    2022

This paper was first reviewed by grok-4.5 on July 31, 2026.