pith. sign in

arxiv: 2606.28397 · v1 · pith:EE3YD7MAnew · submitted 2026-06-24 · 💻 cs.CV · cs.AI

CLOSER-VLN: Closed-Loop Self-Verified Retrieval-Augmented Reasoning for Aerial Vision-Language Navigation

Shaoxuan Li , Xiangyu Dong , Xiaoguang Ma , Junfeng Chen , Haoran Zhao , Yaoming Zhou This is my paper

Pith reviewed 2026-06-30 01:32 UTC · model grok-4.3

classification 💻 cs.CV cs.AI
keywords vision-language navigationaerial navigationclosed-loop reasoningretrieval-augmented reasoningself-verificationCityNav benchmarkerror accumulation
0
0 comments X

The pith

A closed-loop verification-retrieval-correction process improves aerial vision-language navigation to 32.01 percent success on unseen environments without task-specific training.

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

The paper contends that open-loop generation of actions from instructions and observations allows small mistakes to compound into large trajectory errors and lost targets in aerial settings. It introduces CLOSER, a training-free method that runs action reasoning, reliability checks, targeted retrieval from memory, and correction in sequence before any move is executed. The resulting CLOSER-VLN framework is evaluated on the CityNav benchmark and records 32.01 percent success rate and 21.28 percent SPL on the test-unseen split. A reader would care because the approach shows how to make large multimodal models more reliable for navigation by catching and fixing errors internally rather than relying on retraining or external supervision.

Core claim

The paper claims that sequentially executing action reasoning, multidimensional reliability verification, verification-triggered multimodal retrieval, and action correction in a closed loop before concrete execution prevents error accumulation and raises navigation performance in unseen aerial environments.

What carries the argument

The CLOSER loop that chains a hierarchical reasoner, a multidimensional action verifier, and a verification-triggered multimodal retriever which activates only on failed verification.

If this is right

  • Candidate actions are generated, checked, and corrected internally before execution, reducing the chance of irreversible path errors.
  • Retrieval occurs only when verification fails, limiting unnecessary external calls while still supplying targeted exemplars.
  • No task-specific navigation policy is trained, so the same components can be applied across different aerial VLN environments.
  • Performance gains appear on the test-unseen split, indicating better generalization than open-loop baselines.

Where Pith is reading between the lines

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

  • The same verification-triggered retrieval pattern could be inserted into other instruction-following agents where early mistakes compound.
  • Replacing the current reasoner with a stronger multimodal model might further raise the baseline before the loop even activates.
  • Measuring cumulative trajectory deviation at each step would directly test whether the loop reduces error growth as claimed.

Load-bearing premise

Minor errors in intermediate actions accumulate rapidly into large trajectory deviations in aerial VLN, and the verification-retrieval-correction loop can stop this accumulation without any task-specific policy training.

What would settle it

Disabling the verifier and retriever components and measuring whether success rate and SPL on the test-unseen split fall back to open-loop levels, or recording whether trajectory deviation still occurs at the reported rates when the full loop is active.

Figures

Figures reproduced from arXiv: 2606.28397 by Haoran Zhao, Junfeng Chen, Shaoxuan Li, Xiangyu Dong, Xiaoguang Ma, Yaoming Zhou.

Figure 1
Figure 1. Figure 1: Comparison of decision-making methods for aerial vision-language navigation. Zero-shot end-to-end methods (a) and retrieval-augmented methods [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Overall architecture of CLOSER-VLN. The CLOSER-VLN preserves the stage-wise aerial VLN interface inherited from GeoNav, while introducing a [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Overall pipeline of the multidimensional action verifier. Offline [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Verification-triggered multimodal retrieval. The exemplar knowledge base is constructed from successful decision fragments and organized by task [PITH_FULL_IMAGE:figures/full_fig_p007_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Confusion matrices of different verifiers. Rows represent ground [PITH_FULL_IMAGE:figures/full_fig_p010_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: A visualization of the system’s explicit closed-loop reasoning during a UAV navigation episode in the CityNav environment. The CLOSER-VLN takes [PITH_FULL_IMAGE:figures/full_fig_p011_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: A visualization of a verification failure case during a UAV navigation episode in the CityNav environment. The CLOSER-VLN takes the SCM ( [PITH_FULL_IMAGE:figures/full_fig_p011_7.png] view at source ↗
read the original abstract

Vision-language navigation (VLN) has recently advanced with large language and multimodal models, enabling agents to follow natural-language instructions in unseen environments without training a task-specific navigation policy. However, most existing VLN methods relying on large models still adopt an open-loop decision-execution approach, where candidate actions are generated from instructions and observations but are rarely verified or corrected before execution. This causes critical issues in aerial VLN, where minor errors in intermediate actions may quickly accumulate into large trajectory deviations and lead to target loss. To address this issue, we propose Closed-loop Self-verified Retrieval-augmented Reasoning (CLOSER), a training-policy-free method that sequentially performs action reasoning, reliability verification, targeted retrieval, and action correction in a closed-loop manner before executing concrete actions. We instantiate the CLOSER in aerial VLN tasks and develop a CLOSER-VLN framework, which is composed of three components: a hierarchical reasoner for generating candidate actions based on available information, a multidimensional action verifier for assessing the reliability of actions generated by the reasoner, and a verification-triggered multimodal retriever for retrieving targeted exemplars from a memory bank only when verification fails. We conduct experimental evaluations on the CityNav benchmark, where CLOSER-VLN achieves 32.01% SR and 21.28% SPL on the test-unseen split, confirming the effectiveness of closed-loop reasoning.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

0 major / 3 minor

Summary. The paper proposes CLOSER-VLN, a training-free closed-loop framework for aerial vision-language navigation consisting of a hierarchical reasoner that generates candidate actions, a multidimensional action verifier that assesses reliability, and a verification-triggered multimodal retriever that fetches targeted exemplars from a memory bank only on verification failure. The method performs sequential reasoning-verification-retrieval-correction before action execution to mitigate error accumulation. On the CityNav benchmark it reports 32.01% SR and 21.28% SPL on the test-unseen split.

Significance. If the reported numbers are supported by appropriate open-loop baselines, component ablations, and statistical analysis, the work would demonstrate that a verification-retrieval-correction loop can improve long-horizon aerial VLN performance without task-specific policy training. The selective use of retrieval only on verification failure is an efficient design choice that could generalize to other embodied reasoning settings.

minor comments (3)
  1. [Abstract] Abstract: the reported SR and SPL values are presented without any reference to the corresponding open-loop baseline numbers or to the number of evaluation runs, which makes it impossible to judge the magnitude of improvement from the closed-loop components.
  2. [§3] §3 (Method): the description of the multidimensional verifier does not specify the exact dimensions or scoring functions used to assess action reliability; a concrete example or pseudocode would clarify how verification triggers retrieval.
  3. [§4] §4 (Experiments): the paper should include an ablation table isolating the contribution of the verifier and the retriever (e.g., reasoner-only vs. reasoner+verifier vs. full CLOSER-VLN) to substantiate the central claim that the closed loop is responsible for the gains.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for the positive assessment of CLOSER-VLN and the recommendation for minor revision. No major comments were raised.

Circularity Check

0 steps flagged

No significant circularity

full rationale

The paper describes an empirical method (CLOSER-VLN) and reports direct experimental metrics (SR 32.01%, SPL 21.28%) on the CityNav test-unseen split. No derivation chain, equations, or fitted parameters exist that reduce the claimed performance to a self-definition, renamed input, or self-citation load-bearing premise. The framework components are presented as procedural steps whose effectiveness is measured externally on held-out data, satisfying the condition for a self-contained empirical result.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Based solely on the abstract, no free parameters, axioms, or invented entities are explicitly introduced or fitted; the method is described as training-policy-free and relies on existing large models and a memory bank.

pith-pipeline@v0.9.1-grok · 5802 in / 1136 out tokens · 42481 ms · 2026-06-30T01:32:24.595395+00:00 · methodology

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

46 extracted references · 10 canonical work pages · 2 internal anchors

  1. [1]

    Dream to recall: Imagination-guided expe- rience retrieval for memory-persistent vision-and-language navigation,

    Y . Xu, Y . Pan, and Z. Liu, “Dream to recall: Imagination-guided expe- rience retrieval for memory-persistent vision-and-language navigation,” IEEE Transactions on Pattern Analysis and Machine Intelligence, 2026. 13

    2026

  2. [2]

    Etpnav: Evolving topological planning for vision-language navigation in continuous environments,

    D. An, H. Wang, W. Wang, Z. Wang, Y . Huang, K. He, and L. Wang, “Etpnav: Evolving topological planning for vision-language navigation in continuous environments,”IEEE Transactions on Pattern Analysis and Machine Intelligence, 2024

    2024

  3. [3]

    Fine-grained alignment supervision matters in vision-and-language navigation,

    K. He, Y . Huang, Y . Jing, Q. Wu, and L. Wang, “Fine-grained alignment supervision matters in vision-and-language navigation,”IEEE Transac- tions on Pattern Analysis and Machine Intelligence, 2026

    2026

  4. [4]

    Correctable landmark discovery via large models for vision-language navigation,

    B. Lin, Y . Nie, Z. Wei, Y . Zhu, H. Xu, S. Ma, J. Liu, and X. Liang, “Correctable landmark discovery via large models for vision-language navigation,”IEEE Transactions on Pattern Analysis and Machine Intel- ligence, vol. 46, no. 12, pp. 8534–8548, 2024

    2024

  5. [5]

    Evolvenav: Empowering llm-based vision- language navigation via self-improving embodied reasoning,

    B. Lin, Y . Nie, K. L. Zai, Z. Wei, M. Han, R. Xu, M. Niu, J. Han, H. Zhang, L. Linet al., “Evolvenav: Empowering llm-based vision- language navigation via self-improving embodied reasoning,”IEEE Transactions on Pattern Analysis and Machine Intelligence, 2026

    2026

  6. [6]

    Constraint-aware zero-shot vision-language navigation in continuous environments,

    K. Chen, D. An, Y . Huang, R. Xu, Y . Su, Y . Ling, I. Reid, and L. Wang, “Constraint-aware zero-shot vision-language navigation in continuous environments,”IEEE Transactions on Pattern Analysis and Machine Intelligence, 2025

    2025

  7. [7]

    Do as i can, not as i say: Grounding language in robotic affordances,

    M. Ahn, A. Brohan, N. Brown, Y . Chebotar, O. Cortes, B. David, C. Finn, C. Fu, K. Gopalakrishnan, K. Hausmanet al., “Do as i can, not as i say: Grounding language in robotic affordances,” inConference on Robot Learning. PMLR, 2022, pp. 287–318

    2022

  8. [8]

    Imaginenav++: Prompting vision-language models as embodied navigator through scene imagina- tion,

    T. Wang, X. Zhao, W. Cai, and C. Sun, “Imaginenav++: Prompting vision-language models as embodied navigator through scene imagina- tion,”IEEE Transactions on Pattern Analysis and Machine Intelligence, 2026

    2026

  9. [9]

    Navcot: Boosting llm-based vision-and-language navigation via learning disentangled reasoning,

    B. Lin, Y . Nie, Z. Wei, J. Chen, S. Ma, J. Han, H. Xu, X. Chang, and X. Liang, “Navcot: Boosting llm-based vision-and-language navigation via learning disentangled reasoning,”IEEE Transactions on Pattern Analysis and Machine Intelligence, 2025

    2025

  10. [10]

    Flightgpt: Towards generalizable and interpretable uav vision-and-language navigation with vision-language models,

    H. Cai, J. Dong, J. Tan, J. Deng, S. Li, Z. Gao, H. Wang, Z. Su, A. Sumalee, and R. Zhong, “Flightgpt: Towards generalizable and interpretable uav vision-and-language navigation with vision-language models,” inProceedings of the 2025 Conference on Empirical Methods in Natural Language Processing, 2025, pp. 6670–6687

    2025

  11. [11]

    Navgpt: Explicit reasoning in vision- and-language navigation with large language models,

    G. Zhou, Y . Hong, and Q. Wu, “Navgpt: Explicit reasoning in vision- and-language navigation with large language models,” inProceedings of the AAAI Conference on Artificial Intelligence, vol. 38, no. 7, 2024, pp. 7641–7649

    2024

  12. [12]

    Open-nav: Exploring zero-shot vision-and-language navigation in continuous environment with open-source llms,

    Y . Qiao, W. Lyu, H. Wang, Z. Wang, Z. Li, Y . Zhang, M. Tan, and Q. Wu, “Open-nav: Exploring zero-shot vision-and-language navigation in continuous environment with open-source llms,” in2025 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2025, pp. 6710–6717

    2025

  13. [13]

    Inner monologue: Embodied reasoning through planning with language models,

    W. Huang, P. Abbeel, D. Pathak, and I. Mordatch, “Inner monologue: Embodied reasoning through planning with language models,” inCon- ference on Robot Learning. PMLR, 2022, pp. 1769–1782

    2022

  14. [14]

    Rt-2: Vision- language-action models transfer web knowledge to robotic control,

    A. Brohan, N. Brown, J. Carbajal, Y . Chebotar, X. Chen, K. Choro- manski, T. Ding, D. Driess, A. Dubey, C. Finnet al., “Rt-2: Vision- language-action models transfer web knowledge to robotic control,” in Conference on Robot Learning. PMLR, 2023, pp. 2165–2183

    2023

  15. [15]

    Mapgpt: Map-guided prompting with adaptive path planning for vision-and- language navigation,

    J. Chen, B. Lin, R. Xu, Z. Chai, X. Liang, and K.-Y . Wong, “Mapgpt: Map-guided prompting with adaptive path planning for vision-and- language navigation,” inProceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), 2024, pp. 9796–9810

    2024

  16. [16]

    Cmmr-vln: Vision-and- language navigation via continual multimodal memory retrieval,

    H. Li, X. Dong, H. Jiang, Y . Zhou, and X. Ma, “Cmmr-vln: Vision-and- language navigation via continual multimodal memory retrieval,”arXiv preprint arXiv:2603.07997, 2026

    work page arXiv 2026

  17. [17]

    Ragnav: A retrieval-augmented topological reason- ing framework for multi-goal visual-language navigation,

    L. Luo and Q. Bai, “Ragnav: A retrieval-augmented topological reason- ing framework for multi-goal visual-language navigation,”arXiv preprint arXiv:2603.03745, 2026

    work page arXiv 2026

  18. [18]

    Aeri- alvln: Vision-and-language navigation for uavs,

    S. Liu, H. Zhang, Y . Qi, P. Wang, Y . Zhang, and Q. Wu, “Aeri- alvln: Vision-and-language navigation for uavs,” inProceedings of the IEEE/CVF International Conference on Computer Vision, 2023, pp. 15 384–15 394

    2023

  19. [19]

    Citynav: A large-scale dataset for real-world aerial navigation,

    J. Lee, T. Miyanishi, S. Kurita, K. Sakamoto, D. Azuma, Y . Matsuo, and N. Inoue, “Citynav: A large-scale dataset for real-world aerial navigation,” inProceedings of the IEEE/CVF International Conference on Computer Vision, 2025, pp. 5912–5922

    2025

  20. [20]

    Improving language models by retrieving from trillions of tokens,

    S. Borgeaud, A. Mensch, J. Hoffmann, T. Cai, E. Rutherford, K. Milli- can, G. B. van den Driessche, J.-B. Lespiau, B. Damoc, A. Clarket al., “Improving language models by retrieving from trillions of tokens,” in International Conference on Machine Learning. PMLR, 2022, pp. 2206–2240

    2022

  21. [21]

    Rana: Retrieval-augmented navigation,

    G. Monaci, R. S. Rezende, R. Deffayet, G. Csurka, G. Bono, H. D ´ejean, S. Clinchant, and C. Wolf, “Rana: Retrieval-augmented navigation,” arXiv preprint arXiv:2504.03524, 2025

    work page arXiv 2025

  22. [22]

    Exploratory retrieval- augmented planning for continual embodied instruction following,

    M. Yoo, J. Jang, W.-J. Park, and H. Woo, “Exploratory retrieval- augmented planning for continual embodied instruction following,”Ad- vances in Neural Information Processing Systems, vol. 37, pp. 67 034– 67 060, 2024

    2024

  23. [23]

    Geonav: Empowering mllms with dual-scale geospatial reasoning for language- goal aerial navigation,

    H. Xu, Y . Hu, C. Gao, Z. Zhu, Y . Zhao, and Q. Yin, “Geonav: Empowering mllms with dual-scale geospatial reasoning for language- goal aerial navigation,”Pattern Recognition, p. 113365, 2026

    2026

  24. [24]

    Topo- logical planning with transformers for vision-and-language navigation,

    S. Chen, P.-L. Guhur, M. Tapaswi, C. Schmid, and I. Laptev, “Topo- logical planning with transformers for vision-and-language navigation,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2021, pp. 11 276–11 286

    2021

  25. [25]

    Object goal navigation using goal-oriented semantic exploration,

    D. S. Chaplot, D. Gandhi, A. Gupta, and R. Salakhutdinov, “Object goal navigation using goal-oriented semantic exploration,” inAdvances in Neural Information Processing Systems, vol. 33, 2020, pp. 4247– 4258

    2020

  26. [26]

    SE-VLN: A self-evolving vision-language navigation frame- work based on multimodal large language models,

    X. Dong, H. Zhao, J. Gao, H. Li, X. Ma, Y . Zhou, F. Chen, and J. Liu, “Se-vln: A self-evolving vision-language navigation framework based on multimodal large language models,”arXiv preprint arXiv:2507.13152, 2025

    work page arXiv 2025

  27. [27]

    Exploring spatial representation to enhance llm reasoning in aerial vision-language navigation,

    Y . Gao, Z. Wang, P. Han, L. Jing, D. Wang, and B. Zhao, “Exploring spatial representation to enhance llm reasoning in aerial vision-language navigation,”arXiv preprint arXiv:2410.08500, 2024

    work page arXiv 2024

  28. [28]

    Citynavagent: Aerial vision-and-language navigation with hierarchical semantic planning and global memory,

    W. Zhang, C. Gao, S. Yu, R. Peng, B. Zhao, Q. Zhang, J. Cui, X. Chen, and Y . Li, “Citynavagent: Aerial vision-and-language navigation with hierarchical semantic planning and global memory,” inProceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), 2025, pp. 31 292–31 309

    2025

  29. [29]

    Visa-enhanced aerial vln: A visual-spatial reasoning enhanced framework for aerial vision-language navigation,

    H. Tong, X. Dong, X. Ma, H. Zhao, Y . Zhou, and C. Lin, “Visa-enhanced aerial vln: A visual-spatial reasoning enhanced framework for aerial vision-language navigation,”arXiv preprint arXiv:2603.08007, 2026

    work page arXiv 2026

  30. [30]

    Synthesizing scientific literature with retrieval-augmented language models,

    A. Asai, J. He, R. Shao, W. Shi, A. Singh, J. C. Chang, K. Lo, L. Soldaini, S. Feldman, M. D’Arcyet al., “Synthesizing scientific literature with retrieval-augmented language models,”Nature, pp. 1–7, 2026

    2026

  31. [31]

    Embodied large language models enable robots to complete complex tasks in unpredictable environments,

    R. Mon-Williams, G. Li, R. Long, W. Du, and C. G. Lucas, “Embodied large language models enable robots to complete complex tasks in unpredictable environments,”Nature Machine Intelligence, vol. 7, no. 4, pp. 592–601, 2025

    2025

  32. [32]

    A retrieval-augmented framework enabling vlm spatial awareness for object-centric robot manipulation,

    K. Chen, C. Li, C. Tu, J. Pan, Y . Ma, W. Chen, Z. Zhou, X. Xu, S. James, C.-W. Fuet al., “A retrieval-augmented framework enabling vlm spatial awareness for object-centric robot manipulation,”Science Robotics, vol. 11, no. 113, p. eaea2092, 2026

    2026

  33. [33]

    Embodied-RAG: General non-parametric Embodied Memory for Retrieval and Generation

    Q. Xie, S. Y . Min, P. Ji, Y . Yang, T. Zhang, K. Xu, A. Bajaj, R. Salakhutdinov, M. Johnson-Roberson, and Y . Bisk, “Embodied-rag: General non-parametric embodied memory for retrieval and generation,” arXiv preprint arXiv:2409.18313, 2024

    work page arXiv 2024

  34. [34]

    Chain-of-thought prompting elicits reasoning in large language models,

    J. Wei, X. Wang, D. Schuurmans, M. Bosma, B. Ichter, F. Xia, E. H. Chi, Q. V . Le, and D. Zhou, “Chain-of-thought prompting elicits reasoning in large language models,” inAdvances in Neural Information Processing Systems, vol. 35, 2022, pp. 24 824–24 837

    2022

  35. [35]

    Self-consistency improves chain of thought reasoning in language models,

    X. Wang, J. Wei, D. Schuurmans, Q. V . Le, E. H. Chi, S. Narang, A. Chowdhery, and D. Zhou, “Self-consistency improves chain of thought reasoning in language models,” inInternational Conference on Learning Representations, 2023

    2023

  36. [36]

    React: Synergizing reasoning and acting in language models,

    S. Yao, J. Zhao, D. Yu, N. Du, I. Shafran, K. Narasimhan, and Y . Cao, “React: Synergizing reasoning and acting in language models,” in International Conference on Learning Representations, 2023

    2023

  37. [37]

    Tree of thoughts: Deliberate problem solving with large language models,

    S. Yao, D. Yu, J. Zhao, I. Shafran, T. L. Griffiths, Y . Cao, and K. Narasimhan, “Tree of thoughts: Deliberate problem solving with large language models,” inAdvances in Neural Information Processing Systems, vol. 36, 2023

    2023

  38. [38]

    Judging llm-as-a-judge with mt-bench and chatbot arena,

    L. Zheng, W.-L. Chiang, Y . Sheng, S. Zhuang, Z. Wu, Y . Zhuang, Z. Lin, Z. Li, D. Li, E. Xinget al., “Judging llm-as-a-judge with mt-bench and chatbot arena,”Advances in neural information processing systems, vol. 36, pp. 46 595–46 623, 2023

    2023

  39. [39]

    Self-refine: Iter- ative refinement with self-feedback,

    A. Madaan, N. Tandon, P. Gupta, S. Hallinan, L. Gao, S. Wiegreffe, U. Alon, N. Dziri, S. Prabhumoye, Y . Yanget al., “Self-refine: Iter- ative refinement with self-feedback,”Advances in neural information processing systems, vol. 36, pp. 46 534–46 594, 2023

    2023

  40. [40]

    Re- flexion: Language agents with verbal reinforcement learning,

    N. Shinn, F. Cassano, A. Gopinath, K. Narasimhan, and S. Yao, “Re- flexion: Language agents with verbal reinforcement learning,”Advances in neural information processing systems, vol. 36, pp. 8634–8652, 2023

    2023

  41. [41]

    Chain-of-verification reduces hallucination in large 14 language models,

    S. Dhuliawala, M. Komeili, J. Xu, R. Raileanu, X. Li, A. Celikyilmaz, and J. Weston, “Chain-of-verification reduces hallucination in large 14 language models,” inFindings of the association for computational linguistics: ACL 2024, 2024, pp. 3563–3578

    2024

  42. [42]

    Critic: Large language models can self-correct with tool-interactive critiquing,

    Z. Gou, Z. Shao, Y . Gong, Y . Yang, N. Duan, W. Chenet al., “Critic: Large language models can self-correct with tool-interactive critiquing,” inInternational Conference on Learning Representations, vol. 2024, 2024, pp. 57 734–57 811

    2024

  43. [43]

    Augmenting multimodal llms with self-reflective tokens for knowledge- based visual question answering,

    F. Cocchi, N. Moratelli, M. Cornia, L. Baraldi, and R. Cucchiara, “Augmenting multimodal llms with self-reflective tokens for knowledge- based visual question answering,” inProceedings of the Computer Vision and Pattern Recognition Conference, 2025, pp. 9199–9209

    2025

  44. [44]

    Qwen3 Technical Report

    A. Yang, A. Li, B. Yang, B. Zhang, B. Hui, B. Zheng, B. Yu, C. Gao, C. Huang, C. Lvet al., “Qwen3 technical report,”arXiv preprint arXiv:2505.09388, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  45. [45]

    DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models

    Z. Shao, P. Wang, Q. Zhu, R. Xu, J. Song, X. Bi, H. Zhang, M. Zhang, Y . Li, Y . Wuet al., “Deepseekmath: Pushing the limits of mathematical reasoning in open language models,”arXiv preprint arXiv:2402.03300, 2024

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  46. [46]

    Reason-rft: Reinforcement fine-tuning for visual reasoning,

    H. Tan, Y . Ji, X. Hao, X. Chen, P. Wang, Z. Wang, and S. Zhang, “Reason-rft: Reinforcement fine-tuning for visual reasoning of vision language models,”arXiv preprint arXiv:2503.20752, 2025

    work page arXiv 2025