Pith. sign in

REVIEW 3 major objections 5 minor 51 references

MSNav: Zero-Shot Vision-and-Language Navigation with Dynamic Memory and LLM Spatial Reasoning

T0 review · 3 major / 5 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read Zero-shot navigator reaches 50.9% success

desk verdict The headline zero-shot claim on R2R rests on a spatial module fine-tuned on the same Matterport3D data distribution used for evaluation, and the authors concede this for REVERIE but not for R2R; the architecture is plausible and the new dataset has value, but the empirical support is thin. read the letter →

arxiv 2508.16654 v3 pith:SUN2OKKT submitted 2025-08-20 cs.CV

Chenghao Liu , Zhimu Zhou , Jiachen Zhang , Minghao Zhang , Songfang Huang , Huiling Duan This is my paper
classification cs.CV
keywords zero-shotvision-and-languagenavigationmodularagentarchitecturetopologicalmappruningspatialreasoninginstruction-object-spacedatasetsmalllanguagemodelfine-tuningR2RbenchmarkREVERIE
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper claims that zero-shot vision-and-language navigation, normally handled by a single large language model reasoning end-to-end, becomes much more reliable when split into three cooperating modules: a dynamic map memory that prunes stale nodes, a spatial-reasoning module that extracts task-relevant objects and infers destination layouts, and an LLM decision module that chooses actions. To feed the spatial module, the authors construct an instruction-object-space dataset and fine-tune a small language model that outperforms larger commercial models at object-list extraction. On the R2R zero-shot benchmark the full system reaches 50.9% success and 42.6% SPL, ahead of the strongest prior zero-shot LLM planner; on the larger R2R validation-unseen split it reports 46% success, the best zero-shot result. If the findings hold, the modular decomposition and the fine-tuned spatial model offer a practical recipe for improving embodied instruction-following without training a navigation policy.

What carries the argument

The mechanisms: (1) dynamic map pruning, which scores visited nodes by staleness, connectivity, unexplored neighbors, and distance and removes the top-scored ones; (2) the Spatial Module, where a small model fine-tuned on the new Instruction-Object-Space dataset outputs ranked object lists and a destination layout, and an open-vocabulary detector highlights those objects in the best view; (3) an LLM Decision Module that reads instruction, map, layout, and enhanced view and picks the next action. The I-O-S dataset, built from oracle paths and object annotations in the same simulator used for evaluation, powers the fine-tuned spatial model.

What would settle it

Run the R2R zero-shot evaluation with the spatial module replaced by a general-purpose LLM that has never seen the I-O-S data, keeping all other modules identical. If the success rate falls to the level of the map-based baseline, the fine-tuned model's exposure to the evaluation simulator is the source of the claimed gain. An even cleaner test would evaluate MSNav in a different indoor simulator not used to build I-O-S and check whether the 5-point SR advantage persists.

Watch

Extended reading notes

Core claim

MSNav's claim: zero-shot VLN is bottlenecked not by planning but by asking one LLM to remember, ground, and reason at once. The paper separates these into three modules. A Memory Module maintains a topological graph of visited viewpoints and prunes nodes by a weighted score of staleness, connectivity, unexplored neighbors, and distance. A Spatial Module, built from a small model fine-tuned on a new instruction-object-layout dataset, outputs ranked object lists and a predicted destination layout; the object list drives an open-vocabulary detector. A Decision Module LLM reads instruction, map, layout, and observation, and picks the next viewpoint or stop. Results: R2R-ZS SR 50.9/SPL 42.6 (beat

Load-bearing premise

The load-bearing premise is that the fine-tuned spatial model does not simply memorize the test environments: its training data is extracted from oracle paths and object annotations in the same simulator used for evaluation.

Editorial extensions

If this is right

  • If the results hold, decomposing navigation into memory, spatial reasoning, and planning is a practical recipe that works without any learned policy or fine-tuned navigation model.
  • The pruning rule keeps prompt sizes manageable, and on the 25 longest R2R-ZS episodes the paper reports 44% success versus 24% for the map-based baseline with fewer steps.
  • The spatial module can be grafted onto other zero-shot planners: ablations show improvements when only its destination layout is appended to a non-map-based planner, and larger gains when added to a map-based one.
  • A small fine-tuned model at 4B parameters can beat much larger commercial LLMs on object extraction and ranking on the new dataset, suggesting a cheaper route to specialized spatial reasoning.
  • The new instruction-object-space dataset, with over 28,000 samples, may support orthogonal tasks such as instruction grounding and object search.

Reading between the lines

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

  • The I-O-S training set is built from oracle paths and object annotations in the same simulator used for R2R evaluation, so the R2R gain may partly reflect memorization of the evaluation distribution; the paper itself switches to a general-purpose LLM on REVERIE to avoid I-O-S data contamination, and the analogous R2R control would replace the fine-tuned spatial model with a general-purpose LLM.
  • The hand-tuned pruning weights are a clear candidate for learning; the paper's own map-efficiency metric could serve as the training objective.
  • The modular recipe likely transfers to other embodied instruction tasks, such as outdoor delivery or household manipulation, whenever an equivalent instruction-object-layout annotation can be produced.
  • At $0.10-$0.14 per episode, the pruning mechanism keeps per-episode inference cost modest, but whether the accuracy gain is worth that cost at scale is an economic question the paper does not settle.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. MSNav proposes a modular zero-shot vision-and-language navigation framework combining a dynamic topological Memory Module with node pruning, a Spatial Module that extracts task-relevant objects and infers destination layouts, and a Decision Module that uses GPT-4o for action selection. The authors introduce the I-O-S dataset of 28,414 instruction-object-space samples, fine-tune Qwen3-4B into Qwen-Sp, and report state-of-the-art zero-shot results on the R2R subset (SR 50.9, SPL 42.6), the full R2R validation unseen split (SR 46, SPL 40), and a REVERIE subset (SR 45.7, SPL 32.8), as well as improved object-extraction F1/NDCG over GPT-4o on the I-O-S test set. The central claim is that MSNav is the best zero-shot LLM-based VLN system on these benchmarks.

Significance. If the empirical claims hold, the paper makes several useful contributions: the modular decomposition of memory, spatial reasoning, and decision-making is sensible; the dynamic map pruning idea is a meaningful step toward reducing LLM context load; and the I-O-S dataset plus Qwen-Sp provides a concrete resource for spatial instruction understanding. The plug-and-play demonstration of the Spatial Module on NavGPT and MapGPT is also valuable. However, the headline zero-shot comparison is substantially weakened by the use of Qwen-Sp, which is fine-tuned on I-O-S data derived from the same Matterport3D simulation environment used to evaluate R2R. The paper itself acknowledges this contamination risk for REVERIE but not for R2R. In addition, the main numerical claims are based on small evaluation subsets without error bars or significance tests. The significance of the work is therefore conditional on resolving the domain-overlap issue and providing uncertainty quantification.

major comments (3)
  1. [§3.2, Tables 1–2] The R2R experiments are not zero-shot in the same sense as the baselines. Qwen-Sp is LoRA-fine-tuned on the I-O-S dataset, which is constructed from oracle paths and ground-truth object annotations obtained through the same simulator (Matterport3D) used to evaluate R2R. The Decision Module is conditioned on object lists and destination layouts produced by this fine-tuned model. The manuscript explicitly avoids this for REVERIE ('we used GPT-4o in the Spatial Module for a true zero-shot evaluation, avoiding I-O-S data contamination', §4.2), but does not apply the same precaution to R2R, where the largest gains are reported. As a result, the 5.1-point SR improvement over MapGPT in Table 1 may reflect memorization of test-environment regularities rather than generalizable spatial reasoning. The authors should either report the scene-level overlap between I-O-S training and the R2R evaluatio
  2. [§4.2, Tables 1 and 3] The main comparisons are based on small subsets without error bars or significance tests. R2R-ZS has 216 episodes and REVERIE-ZS has 140 episodes. On REVERIE-ZS the SR gap is 4.3 points (45.7 vs 41.4); with 140 episodes and success probabilities around 0.45, the standard error of the difference is roughly 6 percentage points, so the observed difference is not clearly significant at conventional levels. The authors should report bootstrap confidence intervals, paired tests where applicable, or evaluate on the full validation sets. As it stands, the 'state-of-the-art' claim is not supported with the reported precision.
  3. [§3.1, Eq. (1); §4.1, Eq. (4)] The Memory Module's pruning criterion depends on eight free choices: λ_t, λ_d, λ_f, λ_dist, θ_recent-visit, θ_age, N_remove, and t_start. None of these is given a concrete value in the manuscript, and no sensitivity analysis is reported. Similarly, the newly introduced ME metric uses α=0.25 without discussion of its influence. Because dynamic pruning is a core contribution and the comparisons hinge on it, the missing configuration and ablations make the results hard to reproduce and leave open the possibility that these choices were selected on the evaluation sets. Please provide the full hyperparameter setting and an ablation over the most influential parameters.
minor comments (5)
  1. [§4.4] The long-distance analysis selects 25 cases 'exceeding 10 steps' from R2R-ZS. The selection criterion should be specified precisely (which instructions, how the threshold was chosen) to rule out cherry-picking, since the SR difference on this subset is much larger than on the full R2R-ZS set.
  2. [Table 2] The text says MSNav 'surpasses three trained and pretrained methods' but does not name them. Please identify which rows are being referred to, as some trained methods outperform MSNav on SR.
  3. [Figure 2] The caption uses 'Nremove = 1' while Eq. (1) and the text use N_remove. Unify the notation.
  4. [Appendix availability] The paper repeatedly refers to the Appendix for prompt designs, training details, and dataset construction. If the appendix is not part of the submitted manuscript, this limits reproducibility; please state where it is available.
  5. [Throughout] There are minor language issues, e.g., 'enhancing t he agent's ability' in Section 1 and 'we name it as R2R-ZS uniformly' in Section 4.2. A careful proofread is recommended.

Circularity Check

0 steps flagged · score 2.0 of 10

No formal circularity: the I-O-S fine-tuning creates a benchmark-validity limitation, but navigation results are not equivalent to the training objective by construction.

full rationale

The paper's claim chain is empirical rather than formal: MSNav's R2R-ZS success (SR 50.9, SPL 42.6) is obtained by running the full system, not by algebraically propagating a fitted quantity. The Spatial Module's Qwen-Sp is fine-tuned on the I-O-S dataset, built from Matterport3D oracle paths and ground-truth object annotations (Sec. 3.2). Since R2R evaluation runs in the same Matterport3D simulator, there is a genuine distributional overlap between Qwen-Sp's training data and the R2R test episodes. The authors concede this risk when they replace Qwen-Sp with GPT-4o for REVERIE: "we used GPT-4o in the Spatial Module for a true zero-shot evaluation, avoiding I-O-S data contamination" (Sec. 4.2), and in Sec. 3.2: "For zero-shot REVERIE experiments, we avoid direct use of the fine-tuned model, instead leveraging its learned patterns to design prompts for commercial LLMs." This is a real benchmark-validity caveat, but it is not circularity: the object-list and destination-layout outputs are not the same quantity as SR/SPL; the Decision Module is a separately prompted GPT-4o fusing many inputs; and the ablation gains (Table 5) are empirical outcomes rather than consequences of an identity. No load-bearing self-citation, imported uniqueness theorem, or ansatz-smuggling citation appears. The Memory Module's pruning formula (Eq. 1) is a defined heuristic, not fitted to the evaluation metric. The REVERIE run with GPT-4o supplies partially independent evidence, though on a small sample. Overall, no step reduces to its own inputs by construction, so the circularity score is low.

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

The framework relies on several hand-tuned parameters and environment-derived training data; no new physical or conceptual entities are postulated.

free parameters (4)
  • ME penalty alpha = 0.25
    Chosen post hoc in Section 4.1 to balance map coverage and size; no sensitivity analysis.
  • Pruning thresholds theta_recent-visit, theta_age
    Used in Memory Module to select candidate nodes for pruning; values not reported.
  • Pruning weights lambda_t, lambda_d, lambda_f, lambda_dist in Eq. (1)
    Balance staleness, degree, frontier, and distance in pruning priority; no values or tuning procedure given.
  • N_remove and t_start
    Number of nodes pruned per step and step at which pruning begins; not specified in the paper.
assumptions (4)
  • domain assumption Matterport3D simulator provides ground-truth object annotations, bounding boxes, and oracle paths used to build I-O-S and evaluate navigation.
    Section 3.2: I-O-S dataset is derived from oracle paths and ground-truth object annotations; experimental results rely on this simulator.
  • domain assumption YOLO-World open-vocabulary detection correctly labels task-relevant objects in the selected viewpoint image.
    Section 3.2: Visual Input Enhancement uses YOLO-World to annotate objects and guide the agent.
  • domain assumption GPT-4o is a reliable planner and spatial reasoner when given the textual map, layout, and action options.
    Decision Module uses GPT-4o; the prompt's reliability is assumed from prior work.
  • domain assumption The LLM-proposed destination layouts in I-O-S, verified by human annotators, are accurate representations of the scenes.
    Section 3.2: spatial arrangements were generated by an LLM and verified by humans; errors in these layouts propagate to navigation.

how reviews work

0 comments
Cite this review

Pith. "Pith review of MSNav: Zero-Shot Vision-and-Language Navigation with Dynamic Memory and LLM Spatial Reasoning." pith.science (2026). https://pith.science/paper/SUN2OKKT

@misc{pith2026250816654,
  author       = {Pith},
  title        = {Pith review of: MSNav: Zero-Shot Vision-and-Language Navigation with Dynamic Memory and LLM Spatial Reasoning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/SUN2OKKT}},
  note         = {Machine review of arXiv:2508.16654}
}
read the original abstract

Vision-and-Language Navigation (VLN) requires an agent to interpret natural language instructions and navigate complex environments. Current approaches often adopt a "black-box" paradigm, where a single Large Language Model (LLM) makes end-to-end decisions. However, it is plagued by critical vulnerabilities, including poor spatial reasoning, weak cross-modal grounding, and memory overload in long-horizon tasks. To systematically address these issues, we propose Memory Spatial Navigation(MSNav), a framework that fuses three modules into a synergistic architecture, which transforms fragile inference into a robust, integrated intelligence. MSNav integrates three modules: Memory Module, a dynamic map memory module that tackles memory overload through selective node pruning, enhancing long-range exploration; Spatial Module, a module for spatial reasoning and object relationship inference that improves endpoint recognition; and Decision Module, a module using LLM-based path planning to execute robust actions. Powering Spatial Module, we also introduce an Instruction-Object-Space (I-O-S) dataset and fine-tune the Qwen3-4B model into Qwen-Spatial (Qwen-Sp), which outperforms leading commercial LLMs in object list extraction, achieving higher F1 and NDCG scores on the I-O-S test set. Extensive experiments on the Room-to-Room (R2R) and REVERIE datasets demonstrate MSNav's state-of-the-art performance with significant improvements in Success Rate (SR) and Success weighted by Path Length (SPL).

Figures

Figures reproduced from arXiv: 2508.16654 by the authors.

Figure 1
Figure 1. These modules have a mutual collaboration and co [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 1
Figure 1. The MSNav architecture. The agent perceives the environment and then identifies optimal navigable viewpoint(yellow [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Memory Module, illustrating the dynamic construction and pruning of a task-relevant topological map. At step [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figures from the paper (3 more)
Figure 3
Figure 3. Figure 3: Spatial Module architecture depicts the pipeline [PITH_FULL_IMAGE:figures/full_fig_p005_3.png]
Figure 4
Figure 4. Figure 4: 4 Experiments 4.1 Experimental Settings MSNav is evaluated on the R2R (Anderson et al. 2018) and REVERIE (Qi et al. 2020) datasets, which are widely used benchmarks in zero-shot VLN setting, with the I-O-S dataset used to assess Qwen-Sp’s spatial inferring ability. MSN…
Figure 4
Figure 4. Figure 4: MSNav’s iterative VLN process (key steps shown). The Memory Module dynamically prunes environment states, [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

51 extracted references · 38 canonical work pages

  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

    work page 2018

  2. [2]

    D.; and Hitch, G

    Baddeley, A. D.; and Hitch, G. 1974. Working memory. In The Psychology of Learning and Motivation, volume 8, 47--89. Elsevier

    work page 1974

  3. [3]

    M.; Wu, J.; Winter, C.; Hesse, C.; Chen, M.; Sigler, E.; Litwin, M.; Gray, S.; Chess, B.; Clark, J.; Berner, C.; McCandlish, S.; Radford, A.; Sutskever, I.; and Amodei, D

    Brown, T.; Mann, B.; Ryder, N.; Subbiah, M.; Kaplan, J.; Dhariwal, P.; Neelakantan, A.; Shyam, P.; Sastry, G.; Askell, A.; Agarwal, S.; Herbert-Voss, A.; Krueger, G.; Henighan, T.; Child, R.; Ramesh, A.; Ziegler, D. M.; Wu, J.; Winter, C.; Hesse, C.; Chen, M.; Sigler, E.; Litwin, M.; Gray, S.; Chess, B.; Clark, J.; Berner, C.; McCandlish, S.; Radford, A.;...

    work page 2020

  4. [4]

    Chang, A.; Dai, A.; Funkhouser, T.; Halber, M.; Nießner, M.; Savva, M.; Song, S.; Zeng, A.; and Zhang, Y. 2017. Matterport3D: Learning from RGB-D data in indoor environments. In Proceedings of the IEEE International Conference on 3D Vision (3DV)

    work page 2017

  5. [5]

    Chen, H.; Suhr, A.; Misra, D.; Snavely, N.; and Artzi, Y. 2019. Touchdown: Natural language navigation and spatial reasoning in visual street environments. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 12538--12547

    work page 2019

  6. [6]

    Chen, J.; Lin, B.; Xu, R.; Chai, Z.; Liang, X.; and Wong, K.-Y. K. 2024. MapGPT: Map-Guided Prompting with Adaptive Path Planning for Vision-and-Language Navigation. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics

    work page 2024

  7. [7]

    K.; Chuang, J.; Vázquez, M.; and Savarese, S

    Chen, K.; Chen, J. K.; Chuang, J.; Vázquez, M.; and Savarese, S. 2021 a . Topological Planning with Transformers for Vision-and-Language Navigation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 11276--11286

    work page 2021

  8. [8]

    Chen, S.; Guhur, P.-L.; Schmid, C.; and Laptev, I. 2021 b . History Aware Multimodal Transformer for Vision-and-Language Navigation. In Advances in Neural Information Processing Systems (NeurIPS), volume 34, 5834--5847

    work page 2021

Show all 51 references
  1. [9]

    Chen, S.; Guhur, P.-L.; Schmid, C.; and Laptev, I. 2021 c . History aware multimodal transformer for vision-and-language navigation. arXiv preprint arXiv:2104.01814

    2021 arXiv

  2. [10]

    Chen, S.; Guhur, P.-L.; Tapaswi, M.; Schmid, C.; and Laptev, I. 2022 a . Learning from Unlabeled 3D Environments for Vision-and-Language Navigation. arXiv:2208.11781

    2022 arXiv

  3. [11]

    Chen, S.; Guhur, P.-L.; Tapaswi, M.; Schmid, C.; and Laptev, I. 2022 b . Think Global, Act Local: Dual-scale Graph Transformer for Vision-and-Language Navigation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 16537--16547

    2022

  4. [12]

    Chen, S.; Guhur, P.-L.; Tapaswi, M.; Schmid, C.; and Laptev, I. 2022 c . Think Global, Act Local: Dual-scale Graph Transformer for Vision-and-Language Navigation. arXiv:2202.11742

    2022 arXiv

  5. [13]

    Chen, Y.-C.; Li, L.; Yu, L.; El Kholy, A.; Ahmed, F.; Gan, Z.; Cheng, Y.; and Liu, J. 2020. Uniter: Universal image-text representation learning. In European Conference on Computer Vision, 104--120. Springer

    2020

  6. [14]

    Cheng, T.; Song, L.; Ge, Y.; Liu, W.; Wang, X.; and Shan, Y. 2024. YOLO-World: Real-Time Open-Vocabulary Object Detection. arXiv preprint arXiv:2401.17270

    2024 arXiv

  7. [15]

    Deng, Z.; Narasimhan, K.; and Russakovsky, O. 2020. Evolving graphical planner: Contextual global planning for vision-and-language navigation. Advances in Neural Information Processing Systems, 33: 20660--20672

    2020

  8. [16]

    A.; Patai, E

    Epstein, R. A.; Patai, E. Z.; Julian, J. B.; and Spiers, H. J. 2017. The cognitive map in humans: spatial navigation and beyond. Nature Neuroscience, 20(11): 1504--1513

    2017

  9. [17]

    Fried, D.; Hu, R.; Cirik, V.; Rohrbach, A.; Andreas, J.; Morency, L.-P.; Berg-Kirkpatrick, T.; Saenko, K.; Klein, D.; and Darrell, T. 2018. Speaker-follower models for vision-and-language navigation. Advances in Neural Information Processing Systems, 31

    2018

  10. [18]

    Fuentes-Pacheco, J.; Ruiz-Ascencio, J.; and Rendon-Mancha, J. M. 2015. Visual simultaneous localization and mapping: A survey. Artificial Intelligence Review, 43: 55--81

    2015

  11. [19]

    Google DeepMind . 2025. Start Building with Gemini 2.5 Flash. https://developers.googleblog.com/en/start-building-with-gemini-25-flash/. Accessed: 2025-05-19

    2025

  12. [20]

    Guhur, P.-L.; Tapaswi, M.; Chen, S.; Laptev, I.; and Schmid, C. 2021. Airbert: In-domain Pretraining for Vision-and-Language Navigation. arXiv:2108.09105

    2021 arXiv

  13. [21]

    Hao, W.; Li, C.; Li, X.; Carin, L.; and Gao, J. 2020. Towards learning a generic agent for vision-and-language navigation via pre-training. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 13137--13146

    2020

  14. [22]

    Hong, Y.; Wu, Q.; Qi, Y.; Rodriguez-Opazo, C.; and Gould, S. 2021 a . VLN BERT : A Recurrent Vision-and-Language BERT for Navigation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 1643--1653

    2021

  15. [23]

    Hong, Y.; Wu, Q.; Qi, Y.; Rodriguez-Opazo, C.; and Gould, S. 2021 b . VLN BERT: A Recurrent Vision-and-Language BERT for Navigation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 1643--1653

    2021

  16. [24]

    J.; Shen, Y.; Wallis, P.; Allen-Zhu, Z.; Li, Y.; Wang, S.; Wang, L.; and Chen, W

    Hu, E. J.; Shen, Y.; Wallis, P.; Allen-Zhu, Z.; Li, Y.; Wang, S.; Wang, L.; and Chen, W. 2022. LoRA: Low-Rank Adaptation of Large Language Models. In Proceedings of the International Conference on Learning Representations (ICLR)

    2022

  17. [25]

    Huang, H.; Jain, V.; Mehta, H.; Ku, A.; Magalhaes, G.; Baldridge, J.; and Ie, E. 2019. Transferable Representation Learning in Vision-and-Language Navigation. In Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), 7403--7412. IEEE

    2019

  18. [26]

    Ke, L.; Li, X.; Bisk, Y.; Holtzman, A.; Gan, Z.; Liu, J.; Gao, J.; Choi, Y.; and Srinivasa, S. 2019. Tactical Rewind: Self-Correction via Backtracking in Vision-and-Language Navigation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)...

    2019

  19. [27]

    S.; Reid, M.; Matsuo, Y.; and Iwasawa, Y

    Kojima, T.; Gu, S. S.; Reid, M.; Matsuo, Y.; and Iwasawa, Y. 2022. Large Language Models are Zero-Shot Reasoners. arXiv preprint arXiv:2205.11916

    2022 arXiv

  20. [28]

    Krantz, J.; Wijmans, E.; Majumdar, A.; Batra, D.; and Lee, S. 2020. Beyond the nav-graph: Vision-and-language navigation in continuous environments. In European Conference on Computer Vision, 104--120. Springer

    2020

  21. [29]

    Y.; Zhang, P.; Hu, X.; Zhang, L.; Wang, L.; Hu, H.; Dong, L.; Wei, F.; Choi, Y.; and Gao, J

    Li, X.; Yin, X.; Li, C. Y.; Zhang, P.; Hu, X.; Zhang, L.; Wang, L.; Hu, H.; Dong, L.; Wei, F.; Choi, Y.; and Gao, J. 2020. Oscar: Object-Semantics Aligned Pre-training for Vision-Language Tasks. In Proceedings of the European Conference on Computer Vision (ECCV), 121--137. Springer

    2020

  22. [30]

    Lin, B.; Nie, Y.; Wei, Z.; Chen, J.; Ma, S.; Han, J.; Xu, H.; Chang, X.; and Liang, X. 2024. NavCoT: Boosting LLM-Based Vision-and-Language Navigation via Learning Disentangled Reasoning. arXiv preprint arXiv:2403.07376

    2024 arXiv

  23. [31]

    Lin, X.; Li, G.; and Yu, Y. 2021. Scene-Intuitive Agent for Remote Embodied Visual Grounding. arXiv:2103.12944

    2021 arXiv

  24. [32]

    Long, Y.; Li, X.; Cai, W.; and Dong, H. 2023. Discuss Before Moving: Visual Language Navigation via Multi-expert Discussions. arXiv preprint arXiv:2309.11382

    2023 arXiv

  25. [33]

    Ma, C.-Y.; Wu, Z.; AlRegib, G.; Xiong, C.; and Kira, Z. 2019. The Regretful Agent: Heuristic-Aided Navigation Through Progress Estimation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 6732--6740

    2019

  26. [34]

    Malleret, G.; Salin, P.; Mazza, S.; and Plancher, G. 2024. Working memory forgetting: Bridging gaps between human and animal studies. Neuroscience & Biobehavioral Reviews, 163: 105742

    2024

  27. [35]

    OpenAI . 2023. GPT-4 Technical Report. arXiv preprint arXiv:2303.08774

    2023 arXiv

  28. [36]

    OpenAI. 2024. GPT-4o System Card. https://arxiv.org/abs/2410.21276. Accessed: 2025-05-19

    2024 arXiv

  29. [37]

    Pan, B.; Panda, R.; Jin, S.; Feris, R.; Oliva, A.; Isola, P.; and Kim, Y. 2023. LangNav: Language as a Perceptual Representation for Navigation. arXiv preprint arXiv:2310.07889

    2023 arXiv

  30. [38]

    Y.; Shen, C.; and Hengel, A

    Qi, Y.; Wu, Q.; Anderson, P.; Wang, X.; Wang, W. Y.; Shen, C.; and Hengel, A. v. d. 2020. Reverie: Remote embodied visual referring expression in real indoor environments. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 9982--9991

    2020

  31. [39]

    S.; Tripoliti, E.; Marias, K.; Fotiadis, D

    Skaramagkas, V.; Giannakakis, G.; Ktistakis, E.; Manousos, D.; Karatzanis, I.; Tachos, N. S.; Tripoliti, E.; Marias, K.; Fotiadis, D. I.; and Tsiknakis, M. 2023. Review of Eye Tracking Metrics Involved in Emotional and Cognitive Processes. IEEE Reviews in Biomedical Engineerin...

    2023

  32. [40]

    Tan, H.; Yu, L.; and Bansal, M. 2019. Learning to Navigate Unseen Environments: Back Translation with Environmental Dropout. 2610--2621. Minneapolis, Minnesota

    2019

  33. [41]

    Thrun, S. 1998. Learning metric-topological maps for indoor mobile robot navigation. Artificial Intelligence, 99(1): 21--71

    1998

  34. [42]

    Y.; and Zhang, L

    Wang, X.; Huang, Q.; Celikyilmaz, A.; Gao, J.; Shen, D.; Wang, Y.-F.; Wang, W. Y.; and Zhang, L. 2019. Reinforced Cross-Modal Matching and Self-Supervised Imitation Learning for Vision-Language Navigation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Patter...

    2019

  35. [43]

    Wang, Z.; Li, J.; Hong, Y.; Wang, Y.; Wu, Q.; Bansal, M.; Gould, S.; Tan, H.; and Qiao, Y. 2023. Scaling data generation in vision-and-language navigation. In Proceedings of the IEEE/CVF International Conference on Computer Vision, 12009--12020

    2023

  36. [44]

    Wei, J.; Wang, X.; Schuurmans, D.; Bosma, M.; Ichter, B.; Xia, F.; Chi, E.; Le, Q.; and Zhou, D. 2022. Chain-of-thought prompting elicits reasoning in large language models. Advances in Neural Information Processing Systems, 35: 24824--24837

    2022

  37. [45]

    xAI. 2025. Grok 3 Beta — The Age of Reasoning Agents. Accessed: 2025-05-19

    2025

  38. [46]

    Yang, A.; Li, A.; Yang, B.; Zhang, B.; Hui, B.; Zheng, B.; Yu, B.; Gao, C.; Huang, C.; Lv, C.; Zheng, C.; Liu, D.; Zhou, F.; Huang, F.; Hu, F.; Ge, H.; Wei, H.; Lin, H.; Tang, J.; Yang, J.; Tu, J.; Zhang, J.; Yang, J.; Yang, J.; Zhou, J.; Zhou, J.; Lin, J.; Dang, K.; Bao, K.; ...

    2025 arXiv

  39. [47]

    Yao, S.; Zhao, J.; Yu, D.; Du, N.; Shafran, I.; Narasimhan, K.; and Cao, Y. 2022. React: Synergizing reasoning and acting in language models. arXiv preprint arXiv:2210.03629

    2022 arXiv

  40. [48]

    Zhao, Y.; Chen, J.; Gao, C.; Wang, W.; Yang, L.; Ren, H.; Xia, H.; and Liu, S. 2022. Target-Driven Structured Transformer Planner for Vision-Language Navigation. arXiv:2207.11201

    2022 arXiv

  41. [49]

    Zheng, D.; Huang, S.; Zhao, L.; Zhong, Y.; and Wang, L. 2024. Towards Learning a Generalist Model for Embodied Navigation. arXiv:2312.02010

    2024 arXiv

  42. [50]

    E.; and Wu, Q

    Zhou, G.; Hong, Y.; Wang, Z.; Wang, X. E.; and Wu, Q. 2024. NavGPT-2: Unleashing Navigational Reasoning Capability for Large Vision-Language Models. arXiv:2407.12366

    2024 arXiv

  43. [51]

    Zhou, G.; Hong, Y.; and Wu, Q. 2023. NavGPT: Explicit Reasoning in Vision-and-Language Navigation with Large Language Models. arXiv preprint arXiv:2305.16986

    2023 arXiv

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.