Pith. sign in

REVIEW 4 major objections 5 minor 24 references

RoboReflect: A Robotic Reflective Reasoning Framework for Grasping Ambiguous-Condition Objects

T0 review · 4 major / 5 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read A vision-language model can teach a robot to grasp tricky objects by reflecting on its own failed attempts.

desk verdict A plausible self-reflection framework for grasping ambiguous objects, but the evaluation is self-scored and asymmetric, so the claimed improvements are not established. read the letter →

arxiv 2501.09307 v2 pith:XMVCW5A2 submitted 2025-01-16 cs.RO

Zhen Luo , Yixuan Yang , Yanfu Zhang , Feng Zheng This is my paper · ORCID
classification cs.RO
keywords roboticgraspingambiguous-conditionobjectsvision-languagemodelsself-reflectionerrorcorrectionmemorymodulegraspposeestimationautonomousmanipulation
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

RoboReflect is a robotic grasping framework built around the idea that a large vision-language model can act as the robot's own critic and teacher. When a grasp on an ambiguous-condition object such as a half-empty tissue bag, an open-lid cup, or an ice-cream bar fails, the framework prompts the model to explain why and to propose a corrected action, and a second model instance checks that correction before the robot tries again. Successful strategies are stored in a memory dictionary keyed by object description, so later encounters skip the trial-and-error. On eight everyday objects across three categories, the paper reports per-object success rates of 60 to 90 percent, against roughly 0 to 90 percent for the baselines, with average gains of 21.25 points over AnyGrasp, 17.5 points over ReKep, and 50 points over plain GPT-4V. The point is that autonomous reflection and memory can replace human intervention in correcting robotic errors.

What carries the argument

The load-bearing mechanism is the reflective reasoning loop plus its memory. Ambiguous-condition objects are first categorized by the kind of ambiguity they present: deformable soft surfaces, assembled multi-part objects, and objects with protected parts that must not be grasped. On each failure, the self-reflective sub-module wraps the action image, object description, and instructions into a chain-of-thought (step-by-step reasoning) prompt that asks the LVLM to output an error cause $Y$ and an action correction $P$; the discussion sub-module then uses a second LVLM $M_D$ to judge the result $R=(Y,P)$ and revise it if needed, an idea drawn from peer-rating by language models. The corrected strategies are stored in a dictionary whose key is the object description and whose value is the derived understanding, so later tasks can retrieve the strategy directly. Segmentation, depth back-projection, and atomic action APIs all serve to feed this reflection loop and to translate its output into robot motion.

What would settle it

Rerun the eight-object evaluation with an independent physical check of each grasp, for example a force sensor in the gripper or a fixed second camera that confirms the object stays held after the arm lifts, and compare those outcomes with the model's self-reported grasp state; if many attempts the model called successful actually dropped or damaged the object, the reflection and memory gains would not reflect real grasping performance.

Watch

Extended reading notes

Core claim

The paper's central claim is that decomposing a grasp into two verdicts, whether the object was lifted (grasp state $G_S$) and whether the grasp position matched human expectations (grasp position $G_P$), and feeding failed attempts back through a reflective reasoning module lets a robot converge on correct strategies for objects whose condition is ambiguous. The loop is: the action module executes; the judgment module asks GPT-4V whether both $G_S$ and $G_P$ hold; on failure, the self-reflective module produces an error cause $Y$ and a correction $P$ using chain-of-thought reasoning over the object description and action images; and a discussion module with a second LVLM $M_D$ verifies or revises that suggestion before the next attempt. When a trial succeeds, the object description and the derived understanding are stored as a memory entry and reused on future encounters. The reported experiments on eight objects are meant to show that this loop outperforms the AnyGrasp grasp pose estimator, the ReKep relational keypoint planner, and plain GPT-4V planning, and that both the discussion and memory modules contribute to the gain.

Load-bearing premise

The central claim rests on trusting the same vision-language model that plans and reflects to also judge accurately whether its own grasp attempts succeeded and whether the grasp position was appropriate.

Editorial extensions

If this is right

  • Robots using RoboReflect can improve at ambiguous grasping without human feedback, because each failed attempt generates its own corrected strategy.
  • The memory module lifts mixed-task success from about 75 to 80 percent to 90 to 95 percent in the paper's ablation, showing that storing successful strategies is what makes repeated encounters reliable.
  • The discussion module adds an average of 15.2 percentage points of success, with the largest effect on objects that need several reflection rounds, such as cookies and hard drives.
  • Because success is defined jointly by grasp state and grasp position, the framework avoids grasps that lift an object but damage it or touch an unusable part, such as the edible portion of an ice-cream bar.

Reading between the lines

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

  • A natural extension is to apply the same reflect-correct-store loop to other manipulation skills, such as insertion, pouring, or assembly, where failure is visible in the action image and a corrected strategy can be verbalized.
  • The object-keyed memory suggests a continual learning path: a robot could bootstrap knowledge of new object states by analogy to stored entries, so later objects require fewer reflection rounds.
  • A direct test of the framework's robustness would be to replace the LVLM's self-reported grasp judgment with a force/torque sensor or an independent camera check; if physical verification agrees with the model's verdicts, the reported gains stand independently of model self-assessment.
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

4 major / 5 minor

Summary. The paper proposes RoboReflect, a framework that uses GPT-4V as a large vision-language model (LVLM) to perform reflective reasoning for robotic grasping of objects in 'ambiguous conditions.' The system is composed of a visual processing module, an action module, a judgment module that evaluates grasp success and grasp-position correctness, a reflective reasoning module with a self-reflection and a discussion sub-module, and a memory module that stores successful strategies. The authors evaluate on eight everyday objects and report success rates that purportedly outperform the baselines AnyGrasp, ReKep, and a GPT-4V-driven planner. The central claim is that autonomous reflection and memory enable a robot to correct failed grasps without human intervention.

Significance. If substantiated, the framework would be a useful step toward using LVLMs for closed-loop robotic error correction, and the three-category taxonomy of ambiguous-condition objects is a reasonable organizational device. The paper also describes a real-robot setup with eight physical objects, which is a strength over purely simulated studies. However, the current evidence does not establish the central claim: the success metric is produced by the same model whose behavior is being measured, the comparison with baselines is asymmetric, and the reported quantitative improvements are inconsistent with the paper's own tables. These issues are load-bearing because every conclusion about the value of reflection, discussion, and memory rests on the trustworthiness of the success labels and the fairness of the comparison.

major comments (4)
  1. [Sections III-C and IV-A] The evaluation is self-referential. The Judgment Module sends the last action image, instruction, and 3D bounding box to the same LVLM M (GPT-4V) that plans actions and performs reflection, and asks 'Was the robotic arm's grasp successful?' and 'Does the grasping position align with human experience?' Section IV-A further states that the correct grasping position GP is 'always defined by the human experience.' Thus the same model that generates the behavior also decides whether it succeeded. No external ground truth (e.g., force/torque thresholds, object-displacement checks, human-annotated labels, or video adjudication) is reported. Because the central success rates in Table I and the ablation deltas in Tables II and III all depend on these self-generated labels, the reported numbers may reflect the model's leniency or bias rather than actual grasping performance. A concrete fix would be to re-annotate all trials with independent human labels or physical sensors, and to report agreement statistics.
  2. [Section IV-B and Table I] The baseline comparison is asymmetric. RoboReflect is allowed multiple attempts per object—as shown by the parenthetical attempt numbers, e.g., tissue bag (1,2,4) and hard drive (1,2,5,9)—whereas the reported success rates for GPT-4V, AnyGrasp, and ReKep appear to be one-shot success rates. The text in Section IV-C explicitly states that 'the first attempt to grasp each object failed' for RoboReflect. Comparing a multi-attempt system against single-attempt baselines inflates the apparent improvement. To support the claim of superiority, the baselines should be given the same number of retries, or the comparison should be reported on a per-attempt basis with appropriate trial counts.
  3. [Section IV-B, text after Table I] The claimed improvements are inconsistent with the table. Averaging the eight per-object success rates in Table I gives 20.0% for AnyGrasp, 52.5% for GPT-4V, and 18.75% for ReKep as the deltas over RoboReflect, not 21.25%, 50%, and 17.5% as stated. The text should present the exact averages computed from the table, report the number of trials per object and per condition, and provide error bars or confidence intervals. Without trial counts, the percentages in Table I have no stated statistical basis.
  4. [Section IV-A and Table III] No trial counts or variance information are given for any of the success rates, including the memory-module ablation where the authors mention '20 mixed grasps' but report only point percentages. This makes it impossible to assess whether differences such as 75% vs. 90% are meaningful. The authors should report the number of trials per object, per condition, and per ablation arm, together with confidence intervals or a significance test.
minor comments (5)
  1. [Section III-A, Eq. (1)] The notation GS ∪ GP is incorrect for the intended logical conjunction: the condition that a grasp is successful only when both GS and GP hold should be written as GS ∧ GP or GS AND GP, not set union. The surrounding text also says 'GS ∪ GP = 0' when it means 'GS = 0 or GP = 0', which is the negation of the conjunction.
  2. [Section III-D] There is a typo: 'The discussion process primarily involves two steps,, as shown in the Figure 1' has a doubled comma.
  3. [Section IV-C] The text divides objects into 'easy-to-reflect' (six objects listed) and 'hard-to-reflect' and then says 'The remaining three objects require three to four reflection steps.' Since eight objects in total are tested and six are listed as easy, only two remain; the count 'three' is inconsistent.
  4. [Section II and Reference [5]] Reference [5] is given as 'Y AY Robot', but the cited work is 'Yell at Your Robot'; the name should be spelled correctly in the text.
  5. [Figure 3 and Figure 4] The figure captions are minimal and do not explain how the displayed grasp poses correspond to the quantitative success rates; for example, Figure 3 shows only qualitative poses without indicating whether those poses led to successful grasps. Adding per-pose success/failure labels or a link to the table would improve clarity.

Circularity Check

2 steps flagged · score 6.0 of 10

The success metric is self-reported by the same GPT-4V that plans, reflects, and populates memory, so the reported gains are substantially self-referential.

  1. self definitional [Section III-C (Judgment Module) and Section IV-A (Metrics)]
    "In the J M, the last frame image Iact, along with the instruction text prompt Ins and 3D bounding box S, are encapsulated in a prompt to provide the LVLM M with the necessary information to infer the task finished status. ... two additional textual questions are input into the LVLM module: 1) Was the robotic arm's grasp successful? 2) Does the grasping position align with human experience? ... The correct grasping position GP of objects is always defined by the human experience."

    The same LVLM M (GPT-4V) is used in the Action Module to generate actions, in the Judgment Module to label grasp success and grasp position, and in the Reflective Reasoning Module to propose corrections. The reported GS/GP labels are therefore M's subjective answers about M's own actions, with no independent ground truth such as force/torque thresholds, object-displacement checks, or human adjudication. Tables I-III thus measure M judging itself: the success criterion being optimized by the framework is identical to the criterion used to score it, so the claimed outperformance over baselines rests on a closed evaluation loop.

  2. self definitional [Section III-C (storage condition) and Section IV-D (memory ablation)]
    "If the evaluation is deemed successful, i.e. GS ∪ GP = 1, by the J M, the object Oi description information will be stored within the memory module for generalizing to grasp other objects in the future task."

    The memory module is populated only when the same M that plans and reflects declares its own attempt successful. The memory ablation in Section IV-D then compares M-with-memory against M-without-memory using the same self-generated GS/GP labels. The memory benefit is therefore M consulting M's own stored judgments to improve M's own self-judged success scores, with no external verification that the stored strategies correspond to real physical success. This makes the memory improvement claim internal to the model rather than independently validated.

full rationale

The central empirical claim is that RoboReflect outperforms AnyGrasp, ReKep, and GPT-4V on ambiguous-condition grasping. That claim depends entirely on the trustworthiness of the success labels and on the fairness of the comparison, and both are compromised. The judgment module asks the same LVLM M that plans and reflects whether the grasp succeeded and whether the position matches human experience, while Section IV-A defines correct grasp position as always given by human experience; no external ground truth is reported. The memory module is likewise populated only after M declares success, so the memory ablation is a loop within M. This is partial circularity rather than a complete definitional collapse: physical grasps do occur and an external evaluator could in principle relabel the episodes, but as reported the numbers cannot be distinguished from M's self-approval. There is also a comparison-protocol asymmetry, since RoboReflect is allowed multiple attempts per object while the baselines are one-shot and the paper states the first attempt always fails; this is an experimental-design threat to the comparative claim rather than a circular reduction. No self-citation chains or imported uniqueness theorems are present. Score 6 reflects that the central result is substantially self-referential while the framework still contains independent components.

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

The central claim rests on the assumed reliability of GPT-4V as both perception and judge, on the subjective definition of grasp-position correctness, and on the absence of external ground truth. There are no fitted numerical parameters in the method, so the free-parameter ledger is empty.

assumptions (4)
  • domain assumption GPT-4V can reliably detect objects, generate captions and bounding boxes, and enable SAM segmentation from RGB-D input.
    Section III-B assumes the frozen detection head and vision-language capabilities work accurately enough for the grasping pipeline.
  • domain assumption The last-frame action image plus instruction and 3D box allows the LVLM to correctly judge both grasp success and grasp-position correctness.
    Section III-C defines GS and GP through LVLM prompts, with no external or human verification.
  • domain assumption A correct grasp position is definable by general human experience and can be assessed from a single image.
    Sections III-C and IV-A state that correct grasping positions are defined by human experience, but no annotation protocol or human evaluators are used.
  • domain assumption Reflective reasoning with chain-of-thought plus a discussion module produces reliable error corrections without accumulating errors.
    Section III-D motivates the discussion module by citing cumulative CoT errors, but the framework's correction reliability is not independently measured.

how reviews work

0 comments
Cite this review

Pith. "Pith review of RoboReflect: A Robotic Reflective Reasoning Framework for Grasping Ambiguous-Condition Objects." pith.science (2026). https://pith.science/paper/XMVCW5A2

@misc{pith2026250109307,
  author       = {Pith},
  title        = {Pith review of: RoboReflect: A Robotic Reflective Reasoning Framework for Grasping Ambiguous-Condition Objects},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/XMVCW5A2}},
  note         = {Machine review of arXiv:2501.09307}
}
read the original abstract

As robotic technology rapidly develops, robots are being employed in an increasing number of fields. However, due to the complexity of deployment environments or the prevalence of ambiguous-condition objects, the practical application of robotics still faces many challenges, leading to frequent errors. Traditional methods and some LLM-based approaches, although improved, still require substantial human intervention and struggle with autonomous error correction in complex scenarios. In this work, we propose RoboReflect, a novel framework leveraging large vision-language models (LVLMs) to enable self-reflection and autonomous error correction in robotic grasping tasks. RoboReflect allows robots to automatically adjust their strategies based on unsuccessful attempts until successful execution is achieved. The corrected strategies are saved in the memory for future task reference. We evaluate RoboReflect through extensive testing on eight common objects prone to ambiguous conditions of three categories. Our results demonstrate that RoboReflect not only outperforms existing grasp pose estimation methods like AnyGrasp and high-level action planning techniques ReKep with GPT-4V but also significantly enhances the robot's capability to adapt and correct errors independently. These findings underscore the critical importance of autonomous self-reflection in robotic systems while effectively addressing the challenges posed by ambiguous-condition environments.

Figures

Figures reproduced from arXiv: 2501.09307 by the authors.

Figure 1
Figure 1. The RoboReflect Framework for Autonomous Error Correction in Robotic Grasping Tasks. The process begins with the Vision Processing Module, which extracts RGB I and depth images ID, generating 3D spatial positions S. I pass to the LVLM M, which, along with text instructions Ins, generates actions Iact through the Action Module. The Judgment Module evaluates the grasp based on the grasp state GS and grasp position GP … view at source ↗
Figure 2
Figure 2. Objects descriptions. The upper half of tissue bag is often empty due to the exhaustion of materials, which makes them susceptible to deformation. A closed-lid cup has a securely fastened lid. An open-lid cup risks lid-body separation due to loose closure. Cookies are fragile and breakable. The upper half of a hard drive is labeled Untouchable. Sealed cup noodles denote standard intact packaging. Unsealed cup noodle… view at source ↗
Figure 4
Figure 4. [PITH_FULL_IMAGE:figures/full_fig_p005_4.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

24 extracted references · 5 canonical work pages

  1. [1]

    LLM-based Robot Task Planning with Exceptional Handling for General Purpose Service Robots

    R. Wang, Z. Yang, Z. Zhao, X. Tong, Z. Hong, and K. Qian, “Llm-based robot task planning with exceptional handling for general purpose service robots,” arXiv preprint arXiv:2405.15646 , 2024

    work page Pith review arXiv 2024

  2. [2]

    Atom: Leveraging large language models for adaptive task object motion strategies in object rearrangement for service robotics,

    I. Y . Guan, G. Zhang, X. Liu, E. Zhao, and J. Wu, “Atom: Leveraging large language models for adaptive task object motion strategies in object rearrangement for service robotics,” in 2024 10th International Conference on Electrical Engineering, Control and Robotics (EECR) . IEEE, 2024, pp. 8–13

    work page 2024

  3. [3]

    Toward automated programming for robotic assembly using chatgpt,

    A. Macaluso, N. Cote, and S. Chitta, “Toward automated programming for robotic assembly using chatgpt,” arXiv preprint arXiv:2405.08216, 2024

    arXiv 2024

  4. [4]

    Reflect: Summarizing robot ex- periences for failure explanation and correction,

    Z. Liu, A. Bahety, and S. Song, “Reflect: Summarizing robot ex- periences for failure explanation and correction,” arXiv preprint arXiv:2306.15724, 2023

    arXiv 2023

  5. [5]

    Yell at your robot: Improving on-the-fly from language corrections,

    L. X. Shi, Z. Hu, T. Z. Zhao, A. Sharma, K. Pertsch, J. Luo, S. Levine, and C. Finn, “Yell at your robot: Improving on-the-fly from language corrections,” arXiv preprint arXiv:2403.12910 , 2024

    arXiv 2024

  6. [6]

    Rekep: Spatio-temporal reasoning of relational keypoint constraints for robotic manipulation,

    W. Huang, C. Wang, Y . Li, R. Zhang, and L. Fei-Fei, “Rekep: Spatio-temporal reasoning of relational keypoint constraints for robotic manipulation,” arXiv preprint arXiv:2409.01652 , 2024

    arXiv 2024

  7. [7]

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

    H.-S. Fang, C. Wang, H. Fang, M. Gou, J. Liu, H. Yan, W. Liu, Y . Xie, and C. Lu, “Anygrasp: Robust and efficient grasp perception in spatial and temporal domains,” IEEE Transactions on Robotics , 2023

    2023

  8. [8]

    Gpt-4 technical report,

    J. Achiam, S. Adler, S. Agarwal, L. Ahmad, I. Akkaya, F. L. Aleman, D. Almeida, J. Altenschmidt, S. Altman, S. Anadkat, et al. , “Gpt-4 technical report,” arXiv preprint arXiv:2303.08774 , 2023

    arXiv 2023

Show all 24 references
  1. [9]

    Palm: Scaling language modeling with pathways,

    A. Chowdhery, S. Narang, J. Devlin, M. Bosma, G. Mishra, A. Roberts, P. Barham, H. W. Chung, C. Sutton, S. Gehrmann, et al., “Palm: Scaling language modeling with pathways,” Journal of Machine Learning Research , vol. 24, no. 240, pp. 1–113, 2023

    2023

  2. [10]

    Llama: Open and efficient foundation language models,

    H. Touvron, T. Lavril, G. Izacard, X. Martinet, M.-A. Lachaux, T. Lacroix, B. Rozi`ere, N. Goyal, E. Hambro, F. Azhar, et al., “Llama: Open and efficient foundation language models,” arXiv preprint arXiv:2302.13971, 2023

    2023 arXiv

  3. [11]

    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. Hausman, et al. , “Do as i can, not as i say: Grounding language in robotic affordances,” arXiv preprint arXiv:2204.01691, 2022

    2022 arXiv

  4. [12]

    Text2motion: From natural language instructions to feasible plans,

    K. Lin, C. Agia, T. Migimatsu, M. Pavone, and J. Bohg, “Text2motion: From natural language instructions to feasible plans,” Autonomous Robots, vol. 47, no. 8, pp. 1345–1365, 2023

    2023

  5. [13]

    Progprompt: Generating situated robot task plans using large language models,

    I. Singh, V . Blukis, A. Mousavian, A. Goyal, D. Xu, J. Tremblay, D. Fox, J. Thomason, and A. Garg, “Progprompt: Generating situated robot task plans using large language models,” in 2023 IEEE Interna- tional Conference on Robotics and Automation (ICRA) . IEEE, 2023, pp. 11 523–11 530

    2023

  6. [14]

    Code as policies: Language model programs for em- bodied control,

    J. Liang, W. Huang, F. Xia, P. Xu, K. Hausman, B. Ichter, P. Florence, and A. Zeng, “Code as policies: Language model programs for em- bodied control,” in 2023 IEEE International Conference on Robotics and Automation (ICRA) . IEEE, 2023, pp. 9493–9500

    2023

  7. [15]

    Open- vla: An open-source vision-language-action model,

    M. J. Kim, K. Pertsch, S. Karamcheti, T. Xiao, A. Balakrishna, S. Nair, R. Rafailov, E. Foster, G. Lam, P. Sanketi, et al. , “Open- vla: An open-source vision-language-action model,” arXiv preprint arXiv:2406.09246, 2024

    2024 arXiv

  8. [16]

    3d-vla: A 3d vision-language-action generative world model,

    H. Zhen, X. Qiu, P. Chen, J. Yang, X. Yan, Y . Du, Y . Hong, and C. Gan, “3d-vla: A 3d vision-language-action generative world model,” arXiv preprint arXiv:2403.09631, 2024

    2024 arXiv

  9. [17]

    Vision-language foundation models as effective robot imitators,

    X. Li, M. Liu, H. Zhang, C. Yu, J. Xu, H. Wu, C. Cheang, Y . Jing, W. Zhang, H. Liu, et al. , “Vision-language foundation models as effective robot imitators,” arXiv preprint arXiv:2311.01378 , 2023

    2023 arXiv

  10. [18]

    Reflexion: an autonomous agent with dynamic memory and self-reflection. arxiv (2023) doi: 10.48550,

    N. Shinn, B. Labash, and A. Gopinath, “Reflexion: an autonomous agent with dynamic memory and self-reflection. arxiv (2023) doi: 10.48550,” arXiv preprint arxiv.2303.11366, 2023

    2023 arXiv

  11. [19]

    Inner monologue: Embodied reasoning through planning with language models,

    W. Huang, F. Xia, T. Xiao, H. Chan, J. Liang, P. Florence, A. Zeng, J. Tompson, I. Mordatch, Y . Chebotar, et al. , “Inner monologue: Embodied reasoning through planning with language models,” arXiv preprint arXiv:2207.05608, 2022

    2022 arXiv

  12. [20]

    Robots that ask for help: Uncertainty alignment for large language model planners,

    A. Z. Ren, A. Dixit, A. Bodrova, S. Singh, S. Tu, N. Brown, P. Xu, L. Takayama, F. Xia, J. Varley, et al. , “Robots that ask for help: Uncertainty alignment for large language model planners,” arXiv preprint arXiv:2307.01928, 2023

    2023 arXiv

  13. [21]

    Segment anything,

    A. Kirillov, E. Mintun, N. Ravi, H. Mao, C. Rolland, L. Gustafson, T. Xiao, S. Whitehead, A. C. Berg, W.-Y . Lo, et al. , “Segment anything,” in Proceedings of the IEEE/CVF International Conference on Computer Vision , 2023, pp. 4015–4026

    2023

  14. [22]

    Corex: Pushing the boundaries of complex reasoning through multi-model collaboration,

    Q. Sun, Z. Yin, X. Li, Z. Wu, X. Qiu, and L. Kong, “Corex: Pushing the boundaries of complex reasoning through multi-model collaboration,” arXiv preprint arXiv:2310.00280 , 2023

    2023 arXiv

  15. [23]

    Learning to break the loop: Analyzing and mitigating repetitions for neural text generation,

    J. Xu, X. Liu, J. Yan, D. Cai, H. Li, and J. Li, “Learning to break the loop: Analyzing and mitigating repetitions for neural text generation,” Advances in Neural Information Processing Systems , vol. 35, pp. 3082–3095, 2022

    2022

  16. [24]

    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. Xing, et al., “Judging llm-as-a-judge with mt- bench and chatbot arena,” Advances in Neural Information Processing Systems, vol. 36, 2024

    2024

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.