Pith. sign in

REVIEW 4 major objections 7 minor 42 references

Progressive Reasoning with Primitive Correction for Compositional Zero-Shot Learning

T0 review · 4 major / 7 minor · reviewed 2026-07-08 · glm-5.2

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

Pith's one-line read Bidirectional reasoning corrects its own mistakes in zero-shot recognition

desk verdict Bidirectional correction idea is novel for CZSL, but the key ablation (Table 6) does not cleanly support the central claim on the largest benchmark. read the letter →

arxiv 2607.05911 v1 pith:B43EKCDI submitted 2026-07-07 cs.CV

Ziyi Chen , Haoyan Shi , Sunhan Xu , Congyan Lang This is my paper
classification cs.CV
keywords compositionalzero-shotlearningchain-of-thoughtreasoningbidirectionalcorrectionmultimodallargelanguagemodelreinforcementpost-trainingGRPOattribute-objectrecognitionstep-levelrewards
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

The paper proposes PRPC, a framework that recasts compositional zero-shot learning — the task of recognizing unseen attribute-object pairs from known primitives — as a structured five-step chain-of-thought reasoning process inside a multimodal large language model. The central idea is that attribute and object predictions should not be made independently or in a single unidirectional pass, because an early error in one primitive propagates and corrupts the other. Instead, PRPC has the model first predict an object, then an attribute conditioned on that object, then re-correct the object using the attribute as a disambiguating cue, then re-confirm the attribute given the corrected object, and finally emit the composition. This bidirectional correction loop is the paper's core mechanism. The model is trained in two stages: supervised fine-tuning on GPT-4o-generated chain-of-thought traces to learn the five-step format, followed by reinforcement learning post-training using GRPO with step-level exact-match rewards that specifically reinforce correctness at each intermediate reasoning step. Experiments on three CZSL benchmarks (MIT-States, C-GQA, VAW-CZSL) show state-of-the-art results, and controlled error-injection experiments demonstrate that the correction loop can recover correct primitives even when the initial object prediction is deliberately wrong.

What carries the argument

Five-step bidirectional corrective reasoning chain (predict object → predict attribute given object → correct object given attribute → reconfirm attribute given corrected object → output composition), implemented as structured CoT in an MLLM, trained via SFT on GPT-4o-generated traces followed by GRPO-based RL post-training with step-level exact-match rewards

What would settle it

If the bidirectional correction steps (Steps 3–4) were removed and performance remained unchanged or improved, or if the error-injection experiment showed no recovery from deliberately incorrect initial predictions, the central claim that mutual correction reduces error propagation would be undermined.

Watch

Extended reading notes

Core claim

The paper's central finding is that bidirectional corrective reasoning — where attribute predictions refine object predictions and vice versa across structured intermediate steps — systematically outperforms both independent primitive prediction and unidirectional conditional modeling for compositional zero-shot recognition. The authors demonstrate this by reformulating CZSL as an open-form generative reasoning task in an MLLM, constraining the model to a five-step chain where Steps 3 and 4 form a mutual verification loop, and showing that step-level reinforcement learning rewards applied to the correction steps yield the largest gains. The error-injection experiment (Table 3) provides the最直

Load-bearing premise

The entire supervised fine-tuning stage depends on GPT-4o-generated five-step chain-of-thought traces being correct and consistent. The paper states that parsing functions verify correctness but provides no systematic evaluation of the quality, faithfulness, or consistency of these auto-generated reasoning traces. If the generated traces contain systematic errors or hallucinated justifications, the model learns to mimic flawed reasoning patterns.

Editorial extensions

If this is right

  • If bidirectional correction helps for attribute-object composition, the same mutual-correction loop could apply to other structured prediction tasks where sub-decisions are interdependent, such as relation extraction, scene graph generation, or multi-attribute recognition.
  • The open-form generative formulation of CZSL removes the need for a predefined candidate label set at inference, which could make compositional recognition more scalable to open-vocabulary settings — but also makes evaluation harder to standardize against prior closed-set work.
  • The finding that step-level rewards on correction steps (Steps 3–4) outperform uniform or early-step rewards suggests that RL post-training for multi-step reasoning benefits most from rewarding verification behavior rather than initial prediction accuracy.
  • The error-injection recovery result implies that reasoning-based models can partially self-correct from bad initial conditions, a property absent from static similarity-based methods like CLIP.
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 / 7 minor

Summary. This paper proposes PRPC, a Progressive Reasoning framework with Primitive Correction for Compositional Zero-Shot Learning (CZSL). The core idea is to reformulate CZSL as a structured five-step Chain-of-Thought (CoT) reasoning process within a Multimodal Large Language Model (MLLM), where attribute and object predictions mutually correct each other bidirectionally, as opposed to independent prediction or unidirectional conditional modeling. The framework consists of two training stages: Stage I performs supervised fine-tuning (SFT) using GPT-4o-generated CoT traces, and Stage II applies Group Relative Policy Optimization (GRPO) with step-level exact-match rewards to further refine reasoning quality. Experiments are conducted on MIT-States, C-GQA, and VAW-CZSL under two evaluation settings (open-form generation and CLIP-similarity-based closed-set classification), with ablations on training stages, CoT designs, and reward weightings.

Significance. The paper introduces a novel reasoning-centric perspective to CZSL, a domain traditionally dominated by embedding-based and disentanglement-based approaches. The formulation of compositional recognition as bidirectional corrective reasoning in MLLMs is a conceptually interesting contribution. The two-stage training pipeline (SFT + GRPO with step-level rewards) is technically detailed, and the error injection experiments (Table 3) provide a falsifiable test of the correction mechanism's robustness. The approach is evaluated on three standard benchmarks with multiple MLLM baselines, and the qualitative examples (Table 5) illustrate the correction process concretely.

major comments (4)
  1. Table 6, MIT-States row: The central claim of the paper is that bidirectional corrective reasoning (Steps 3–5) outperforms unidirectional conditional modeling (Steps 1–2 only) for compositional generalization. However, the ablation directly contradicts this on the largest benchmark. 'PRPC w/o Steps3-5' achieves HM=26.7 and Unseen=28.5, while the full bidirectional model achieves HM=26.4 and Unseen=25.0. The correction steps thus decrease harmonic mean by 0.3 points and unseen accuracy by 3.5 points on MIT-States. The full model only improves Seen accuracy (27.4 vs. 25.2), suggesting overfitting to seen compositions rather than improving zero-shot generalization. This is the most direct test of the paper's core contribution, and it does not support the central claim on the primary benchmark. The text in §4.3 states 'The results clearly demonstrate the effectiveness of using CoT for Bidir[
  2. Table 6: VAW-CZSL results are entirely omitted from the bidirectional correction ablation, leaving only C-GQA showing a benefit (HM 20.8→21.9, modest). Without VAW-CZSL, the ablation covers 2 of 3 benchmarks, and on 1 of those 2 the central claim is contradicted. The authors should include VAW-CZSL in Table 6 to allow readers to assess whether bidirectional correction generalizes.
  3. Table 4, MIT-States and VAW-CZSL rows: Stage II (RL) alone sometimes degrades performance relative to Stage I (SFT). On MIT-States, AUC drops from 11.5 (Stage I) to 10.8 (Stage II), and on VAW-CZSL from 1.0 to 0.7. The paper's final PRPC results (combining both stages) show gains, but the claim that 'GRPO post-training enhances reliability and logical consistency' (§1, §3.6) is undermined when Stage II alone consistently underperforms Stage I. The authors attribute this to sparse rewards and malformed trajectories (§4.3), but this explanation is not rigorously tested. A more thorough analysis of why RL degrades performance on some benchmarks — and whether the gains in the full model come from SFT or RL — is needed.
  4. §3.4: The entire SFT pipeline depends on GPT-4o-generated CoT traces, but no systematic quality evaluation of the generated data is provided. The paper states only that 'parsing functions verify the correctness.' Parsing functions can verify format and exact-match of the final label, but they cannot assess whether the intermediate reasoning steps are faithful, non-hallucinated, or logically consistent. If the generated traces contain systematic reasoning errors, the model learns to mimic flawed patterns. A human or automated quality assessment of a sample of generated CoT traces (e.g., step-level accuracy, logical consistency rate) would substantially strengthen the contribution.
minor comments (7)
  1. Table 2 and Table 3: The column headers are ambiguous. 'Obj-r' and 'Attr-r' appear twice with different values. It is unclear which columns correspond to which step. Clearer labeling (e.g., 'Obj-r (Step 1)', 'Obj-r (Step 3)') would improve readability.
  2. §3.1: The term 'open-form compositional prediction' is introduced but not clearly contrasted with standard 'open-world CZSL.' The relationship between these settings should be clarified.
  3. Table 1: The caption mentions 'Best accuracy values of primitives {Attr, Obj} are also reported,' but it is unclear whether these are from Setting 1 or Setting 2. This should be specified.
  4. §4.2: The paper claims PRPC 'achieves state-of-the-art performance' but Table 1 shows CLIP outperforming PRPC on AUC on MIT-States (11.0 vs. 11.5 — a marginal difference) and VAW-CZSL (0.2 vs. 1.4). The SOTA claim should be qualified more carefully, especially given that PRPC uses the same CLIP text encoder for Setting 2.
  5. Figure 2: The system prompt text is too small to read clearly. A higher-resolution or simplified version would help.
  6. The paper uses 'Qwen3.0-VL-8B' in Table 1 but cites 'Qwen-VL [Bai et al., 2025]' whose title refers to 'Qwen2.5-VL.' The version should be consistently referenced.
  7. §3.6, Eq. (5): The condition 'w_k > 0 only if both r_ans and r_fmt equal 1' is described in text but not reflected in the equation. This conditional gating should be made explicit in the formulation.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the careful and constructive review. The referee raises four major points: (1) the bidirectional correction ablation in Table 6 appears to contradict the central claim on MIT-States; (2) VAW-CZSL is missing from that ablation; (3) Stage II (RL) alone degrades performance on some benchmarks, undermining the claim about GRPO; and (4) no systematic quality evaluation of GPT-4o-generated CoT traces is provided. We address each point below.

read point-by-point responses

  1. Referee: Table 6, MIT-States row: The central claim of the paper is that bidirectional corrective reasoning (Steps 3–5) outperforms unidirectional conditional modeling (Steps 1–2 only) for compositional generalization. However, the ablation directly contradicts this on the largest benchmark. 'PRPC w/o Steps3-5' achieves HM=26.7 and Unseen=28.5, while the full bidirectional model achieves HM=26.4 and Unseen=25.0. The correction steps thus decrease harmonic mean by 0.3 points and unseen accuracy by 3.5 points on MIT-States. The full model only improves Seen accuracy (27.4 vs. 25.2), suggesting overfitting to seen compositions rather than improving zero-shot generalization. This is the most direct test of the paper's core contribution, and it does not support the central claim on the primary benchmark. The text in §4.3 states 'The results clearly demonstrate the effectiveness of using CoT for Bidir[

    Authors: The referee is correct that on MIT-States under the Setting 1 (open-form generation) evaluation, the variant without Steps 3–5 achieves slightly higher HM (26.7 vs. 26.4) and Unseen (28.5 vs. 25.0) than the full five-step model. We acknowledge that this is a genuine weakness in the current manuscript and that the text in §4.3 overstates the evidence on this benchmark. However, we respectfully disagree that this invalidates the core contribution, for three reasons. First, Table 6 evaluates only Stage I (SFT) variants — none of the rows include Stage II (RL). The full PRPC model (SFT + RL) in Table 4 achieves HM=29.2 and Unseen=30.7 on MIT-States, substantially outperforming both Stage I ablation variants. The benefit of bidirectional correction is realized most strongly when combined with RL post-training, because the step-level rewards in Stage II specifically target the correction steps (Steps 3–4) and the reward weighting ablation in Table 7 shows that higher rewards on later steps yield the best performance. Second, the error injection experiment (Table 3) provides a controlled, falsifiable test: when an incorrect object is injected at Step 1, the correction steps recover accuracy (e.g., on MIT-States, Pair-r improves from 1.0 to 12.7). This directly demonstrates that the correction mechanism functions as intended. Third, on C-GQA the ablation does show a consistent benefit (HM 20.8→21.9, Unseen 19.3→20.6). We concede that the MIT-States Setting 1 result is mixed and that the current text is too strong. In the revision we will (a) revise the claim in §4.3 to accurately reflect that bidirectional correction shows clear benefits on C-GQA and in the controlled error-injection setting, but yields mixed results on MIT-States under Setting 1 alone; (b) add a discussion of当 revision: no

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity found — derivation is self-contained with no self-citations

full rationale

The paper's central methodological contribution (the five-step bidirectional corrective reasoning template) is an independent design choice, not derived from or equivalent to a prior result by the authors. The GRPO objective (Eq. 8) is adapted from DeepSeek-R1 [Guo et al., 2025] with proper external attribution. Step-wise rewards (Eq. 4–5) are defined by exact matching against ground-truth labels, not fitted parameters renamed as predictions. Scanning all references, there are zero self-citations — the authors cite only external work (DeepSeek-R1, CoT prompting, Qwen-VL, GPT-4o). No uniqueness theorem is invoked. The GPT-4o CoT data generation pipeline (§3.4) is a data-quality concern (correctness risk), not a circularity issue. The Table 6 ablation honestly reports that bidirectional correction does not improve HM on MIT-States, which is an empirical support concern but is the opposite of circular — it is a direct, non-circular test of the central claim. The derivation chain is self-contained against external benchmarks.

Assumptions & free parameters 8 free parameters · 4 assumptions · 0 invented entities

The paper introduces no new entities (particles, forces, dimensions, etc.). It is a machine learning method paper. The free parameters are all hyperparameters of the training objective that are not reported in the paper. The axioms are domain assumptions and design choices specific to this work.

free parameters (8)
  • α (CoT token loss weight)
    Weight on reasoning token loss in Eq. 2; value not reported in the paper.
  • β (answer token loss weight)
    Weight on answer token loss in Eq. 2; value not reported.
  • w_ans (answer accuracy reward weight)
    Weight on final answer correctness in Eq. 5; value not reported.
  • w_fmt (format reward weight)
    Weight on format compliance in Eq. 5; value not reported.
  • w_k (step-wise reward weights, k=1..5)
    Per-step reward weights in Eq. 5; Table 7 shows 'higher rewards in later steps' is best but exact values not reported.
  • δ (PPO clip range)
    Clip range in GRPO objective (Eq. 8); value not reported.
  • β_KL (KL penalty coefficient)
    KL penalty coefficient in Eq. 8-9; value not reported.
  • G (group size for GRPO sampling)
    Number of trajectories sampled per input in GRPO; value not reported.
assumptions (4)
  • domain assumption Objects dominate visual representations, justifying predicting the object first in Step 1.
    Stated in §3.3: 'Since objects typically dominate visual representations [Wang et al., 2023b], Step1 queries the model for the object category first.' This is a domain assumption borrowed from prior CZSL work.
  • ad hoc to paper GPT-4o can generate faithful, high-quality chain-of-thought reasoning traces for CZSL images when given ground-truth labels.
    §3.4 relies entirely on GPT-4o to generate the five-step CoT supervision. No systematic quality evaluation of generated traces is provided beyond parsing verification.
  • ad hoc to paper Exact-match step-wise rewards provide meaningful supervision for intermediate reasoning quality.
    §3.6 defines rewards as binary exact-match (Eq. 4). This assumes that matching the ground-truth step string indicates correct reasoning, which may not hold for free-form text generation.
  • ad hoc to paper The five-step reasoning template captures the essential structure of compositional recognition.
    §3.3 fixes the reasoning structure as: object → attribute → correct object → reconfirm attribute → conclude. No justification is given for why exactly five steps, or why this particular ordering, beyond intuition.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Progressive Reasoning with Primitive Correction for Compositional Zero-Shot Learning." pith.science (2026). https://pith.science/paper/B43EKCDI

@misc{pith2026260705911,
  author       = {Pith},
  title        = {Pith review of: Progressive Reasoning with Primitive Correction for Compositional Zero-Shot Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/B43EKCDI}},
  note         = {Machine review of arXiv:2607.05911}
}
read the original abstract

Compositional Zero-Shot Learning (CZSL) aims to combine known attributes and objects as primitives for recognizing previously unseen attribute-object pairs. Prior works either predict attributes and objects independently, missing their strong contextual dependency, or use unidirectional conditional modeling (e.g., object-guided attribute prediction), which is prone to error propagation. We propose PRPC, a Progressive Reasoning framework with Primitive Correction, which explicitly models the bidirectional dependency between attributes and objects via step-wise inference. PRPC performs mutual correction of primitives to suppress prediction errors in earlier steps. Specifically, we formulate CZSL as structured, Q&A-style Chain-of-Thought reasoning process and constrain the MLLM to follow predefined semantic steps to generate intermediate decisions. To further enhance the reliability and logical consistency of intermediate reasoning, we introduce reinforcement learning post-training with a GRPO-based objective, providing step-level rewards aligned with the progressive inference procedure. Extensive experiments on three CZSL benchmarks demonstrate that PRPC achieves state-of-the-art performance, validating the effectiveness of progressive reasoning and bidirectional correction for robust compositional generalization.

Figures

Figures reproduced from arXiv: 2607.05911 by the authors.

Figure 1
Figure 1. Comparison of compositional learning paradigms. (a) [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Illustration of our framework PRPC with two-stage training. During training, PRPC uses a CoT-style output format to support step [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

42 extracted references · 42 canonical work pages

  1. [1]

    Qwen2.5-VL Technical Report

    [Baiet al., 2025 ] Shuai Bai, Keqin Chen, Xuejing Liu, Jialin Wang, Wenbin Ge, Sibo Song, Kai Dang, Peng Wang, Shi- jie Wang, Jun Tang, et al. Qwen2.5-vl technical report. arXiv preprint arXiv:2502.13923,

    work page Pith review arXiv 2025

  2. [2]

    Prompting language-informed distribution for compositional zero-shot learning

    [Baoet al., 2024 ] Wentao Bao, Lichang Chen, Heng Huang, and Yu Kong. Prompting language-informed distribution for compositional zero-shot learning. InECCV, pages 107–123. Springer,

    work page 2024

  3. [3]

    Event-centric multi-modal fusion method for dense video captioning.Neural Net- works, 146:120–129,

    [Changet al., 2022 ] Zhi Chang, Dexin Zhao, Huilin Chen, Jingdan Li, and Pengfei Liu. Event-centric multi-modal fusion method for dense video captioning.Neural Net- works, 146:120–129,

    work page 2022

  4. [4]

    Expanding Performance Boundaries of Open-Source Multimodal Models with Model, Data, and Test-Time Scaling

    [Chenet al., 2024 ] Zhe Chen, Weiyun Wang, Yue Cao, Yangzhou Liu, Zhangwei Gao, Erfei Cui, Jinguo Zhu, Shenglong Ye, Hao Tian, Zhaoyang Liu, et al. Expanding performance boundaries of open-source multimodal mod- els with model, data, and test-time scaling.arXiv preprint arXiv:2412.05271,

    work page Pith review arXiv 2024

  5. [5]

    Deep reinforcement learning from human preferences

    [Christianoet al., 2017 ] Paul F Christiano, Jan Leike, Tom Brown, Miljan Martic, Shane Legg, and Dario Amodei. Deep reinforcement learning from human preferences. NeurIPS, 30,

    work page 2017

  6. [6]

    Instructblip: Towards general-purpose vision-language models with instruction tuning.NeurIPS, 36:49250–49267,

    [Daiet al., 2023 ] Wenliang Dai, Junnan Li, Dongxu Li, An- thony Tiong, Junqi Zhao, Weisheng Wang, Boyang Li, Pascale N Fung, and Steven Hoi. Instructblip: Towards general-purpose vision-language models with instruction tuning.NeurIPS, 36:49250–49267,

    work page 2023

  7. [7]

    Molmo and pixmo: Open weights and open data for state-of-the-art multimodal models.arXiv e- prints, pages arXiv–2409,

    [Deitkeet al., 2024 ] Matt Deitke, Christopher Clark, Sangho Lee, Rohun Tripathi, Yue Yang, Jae Sung Park, Moham- madreza Salehi, Niklas Muennighoff, Kyle Lo, Luca Sol- daini, et al. Molmo and pixmo: Open weights and open data for state-of-the-art multimodal models.arXiv e- prints, pages arXiv–2409,

    work page 2024

  8. [8]

    DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning

    [Guoet al., 2025 ] Daya Guo, Dejian Yang, Haowei Zhang, Junxiao Song, Ruoyu Zhang, Runxin Xu, Qihao Zhu, Shi- rong Ma, Peiyi Wang, Xiao Bi, et al. Deepseek-r1: In- centivizing reasoning capability in llms via reinforcement learning.arXiv preprint arXiv:2501.12948,

    work page Pith review arXiv 2025

Show all 42 references
  1. [9]

    Visual programming: Compositional visual reasoning without training

    [Gupta and Kembhavi, 2023] Tanmay Gupta and Aniruddha Kembhavi. Visual programming: Compositional visual reasoning without training. InCVPR, pages 14953–14962,

    2023

  2. [10]

    Learning attention as disentangler for composi- tional zero-shot learning

    [Haoet al., 2023 ] Shaozhe Hao, Kai Han, and Kwan-Yee K Wong. Learning attention as disentangler for composi- tional zero-shot learning. InCVPR, pages 15315–15324,

    2023

  3. [11]

    Troika: Multi-path cross-modal traction for compositional zero- shot learning

    [Huanget al., 2024 ] Siteng Huang, Biao Gong, Yutong Feng, Min Zhang, Yiliang Lv, and Donglin Wang. Troika: Multi-path cross-modal traction for compositional zero- shot learning. InCVPR, pages 24005–24014,

    2024

  4. [12]

    Gqa: A new dataset for real-world vi- sual reasoning and compositional question answering

    [Hudson and Manning, 2019] Drew A Hudson and Christo- pher D Manning. Gqa: A new dataset for real-world vi- sual reasoning and compositional question answering. In CVPR, pages 6700–6709,

    2019

  5. [13]

    Gpt-4o system card.arXiv preprint arXiv:2410.21276,

    [Hurstet al., 2024 ] Aaron Hurst, Adam Lerer, Adam P Goucher, Adam Perelman, Aditya Ramesh, Aidan Clark, AJ Ostrow, Akila Welihinda, Alan Hayes, Alec Rad- ford, et al. Gpt-4o system card.arXiv preprint arXiv:2410.21276,

    2024 arXiv

  6. [14]

    Discovering states and transformations in image collections

    [Isolaet al., 2015 ] Phillip Isola, Joseph J Lim, and Edward H Adelson. Discovering states and transformations in image collections. InCVPR, pages 1383–1391,

    2015

  7. [15]

    Mdetr-modulated detection for end-to-end multi- modal understanding

    [Kamathet al., 2021 ] Aishwarya Kamath, Mannat Singh, Yann LeCun, Gabriel Synnaeve, Ishan Misra, and Nicolas Carion. Mdetr-modulated detection for end-to-end multi- modal understanding. InICCV, pages 1780–1790,

    2021

  8. [16]

    Hierarchical visual primi- tive experts for compositional zero-shot learning

    [Kimet al., 2023 ] Hanjae Kim, Jiyoung Lee, Seongheon Park, and Kwanghoon Sohn. Hierarchical visual primi- tive experts for compositional zero-shot learning. InICCV, pages 5675–5685,

    2023

  9. [17]

    Visual instruction tuning.NeurIPS, 36:34892–34916,

    [Liuet al., 2023 ] Haotian Liu, Chunyuan Li, Qingyang Wu, and Yong Jae Lee. Visual instruction tuning.NeurIPS, 36:34892–34916,

    2023

  10. [18]

    Llavanext: Improved reasoning, ocr, and world knowl- edge,

    [Liuet al., 2024 ] Haotian Liu, Chunyuan Li, Yuheng Li, Bo Li, Yuanhan Zhang, Sheng Shen, and Yong Jae Lee. Llavanext: Improved reasoning, ocr, and world knowl- edge,

    2024

  11. [19]

    From red wine to red tomato: Composition with context

    [Misraet al., 2017 ] Ishan Misra, Abhinav Gupta, and Mar- tial Hebert. From red wine to red tomato: Composition with context. InCVPR, pages 1792–1801,

    2017

  12. [20]

    Learning graph embeddings for compositional zero-shot learning

    [Naeemet al., 2021 ] Muhammad Ferjad Naeem, Yongqin Xian, Federico Tombari, and Zeynep Akata. Learning graph embeddings for compositional zero-shot learning. In CVPR, pages 953–962,

    2021

  13. [21]

    Attributes as operators: factorizing unseen attribute-object compositions

    [Nagarajan and Grauman, 2018] Tushar Nagarajan and Kris- ten Grauman. Attributes as operators: factorizing unseen attribute-object compositions. InECCV, pages 169–185,

    2018

  14. [22]

    Learning to compose soft prompts for composi- tional zero-shot learning

    [Nayaket al., 2023 ] Nihal V Nayak, Peilin Yu, and Stephen Bach. Learning to compose soft prompts for composi- tional zero-shot learning. InICLR,

    2023

  15. [23]

    Training language models to follow instruc- tions with human feedback.NeurIPS, 35:27730–27744,

    [Ouyanget al., 2022 ] Long Ouyang, Jeffrey Wu, Xu Jiang, Diogo Almeida, Carroll Wainwright, Pamela Mishkin, Chong Zhang, Sandhini Agarwal, Katarina Slama, Alex Ray, et al. Training language models to follow instruc- tions with human feedback.NeurIPS, 35:27730–27744,

    2022

  16. [24]

    Learning to predict visual attributes in the wild

    [Phamet al., 2021 ] Khoi Pham, Kushal Kafle, Zhe Lin, Zhi- hong Ding, Scott Cohen, Quan Tran, and Abhinav Shri- vastava. Learning to predict visual attributes in the wild. InCVPR, pages 13018–13028,

    2021

  17. [25]

    What does a platypus look like? generating customized prompts for zero-shot image classification

    [Prattet al., 2023 ] Sarah Pratt, Ian Covert, Rosanne Liu, and Ali Farhadi. What does a platypus look like? generating customized prompts for zero-shot image classification. In ICCV, pages 15691–15701,

    2023

  18. [26]

    Task-driven modular networks for zero-shot compositional learning

    [Purushwalkamet al., 2019 ] Senthil Purushwalkam, Maxi- milian Nickel, Abhinav Gupta, and Marc’Aurelio Ranzato. Task-driven modular networks for zero-shot compositional learning. InICCV, pages 3593–3602,

    2019

  19. [27]

    Learning clustering-based prototypes for compositional zero-shot learning

    [Quet al., 2025 ] Hongyu Qu, Jianan Wei, Xiangbo Shu, and Wenguan Wang. Learning clustering-based prototypes for compositional zero-shot learning. InICLR,

    2025

  20. [28]

    Learning transferable visual models from nat- ural language supervision

    [Radfordet al., 2021 ] Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agar- wal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, et al. Learning transferable visual models from nat- ural language supervision. InICML, pages 8748–8763,

    2021

  21. [29]

    Independent prototype propagation for zero- shot compositionality.NeurIPS, 34:10641–10653,

    [Ruiset al., 2021 ] Frank Ruis, Gertjan Burghouts, and Doina Bucur. Independent prototype propagation for zero- shot compositionality.NeurIPS, 34:10641–10653,

    2021

  22. [30]

    Disentangling visual embeddings for at- tributes and objects

    [Sainiet al., 2022 ] Nirat Saini, Khoi Pham, and Abhinav Shrivastava. Disentangling visual embeddings for at- tributes and objects. InCVPR, pages 13658–13667,

    2022

  23. [31]

    Beyond seen primitive concepts and attribute- object compositional learning

    [Sainiet al., 2024 ] Nirat Saini, Khoi Pham, and Abhinav Shrivastava. Beyond seen primitive concepts and attribute- object compositional learning. InCVPR, pages 14466– 14476,

    2024

  24. [32]

    Aligning large multimodal models with factually aug- mented rlhf

    [Sunet al., 2024 ] Zhiqing Sun, Sheng Shen, Shengcao Cao, Haotian Liu, Chunyuan Li, Yikang Shen, Chuang Gan, Liangyan Gui, Yu-Xiong Wang, Yiming Yang, et al. Aligning large multimodal models with factually aug- mented rlhf. InACL, pages 13088–13110,

    2024

  25. [33]

    Vipergpt: Visual inference via python execu- tion for reasoning

    [Sur´ıset al., 2023] D´ıdac Sur´ıs, Sachit Menon, and Carl V ondrick. Vipergpt: Visual inference via python execu- tion for reasoning. InICCV, pages 11888–11898,

    2023

  26. [34]

    Chain-of-thought prompting elicits reasoning in large language models.NeurIPS, 35:24824– 24837,

    [Weiet al., 2022 ] Jason Wei, Xuezhi Wang, Dale Schuur- mans, Maarten Bosma, Fei Xia, Ed Chi, Quoc V Le, Denny Zhou, et al. Chain-of-thought prompting elicits reasoning in large language models.NeurIPS, 35:24824– 24837,

    2022

  27. [35]

    Deepseek-vl2: Mixture-of-experts vision-language mod- els for advanced multimodal understanding.arXiv preprint arXiv:2412.10302,

    [Wuet al., 2024 ] Zhiyu Wu, Xiaokang Chen, Zizheng Pan, Xingchao Liu, Wen Liu, Damai Dai, Huazuo Gao, Yiyang Ma, Chengyue Wu, Bingxuan Wang, et al. Deepseek-vl2: Mixture-of-experts vision-language mod- els for advanced multimodal understanding.arXiv preprint arXiv:2412.10302,

    2024 arXiv

  28. [36]

    Relation-aware compo- sitional zero-shot learning for attribute-object pair recog- nition.IEEE TMM, 24:3652–3664,

    [Xuet al., 2021 ] Ziwei Xu, Guangzhi Wang, Yongkang Wong, and Mohan S Kankanhalli. Relation-aware compo- sitional zero-shot learning for attribute-object pair recog- nition.IEEE TMM, 24:3652–3664,

    2021

  29. [37]

    Vigor: Im- proving visual grounding of large vision language models with fine-grained reward modeling

    [Yanet al., 2024 ] Siming Yan, Min Bai, Weifeng Chen, Xiong Zhou, Qixing Huang, and Li Erran Li. Vigor: Im- proving visual grounding of large vision language models with fine-grained reward modeling. InECCV, pages 37–

    2024

  30. [38]

    Tree of thoughts: Deliberate problem solving with large language models.NeurIPS, 36:11809–11822,

    [Yaoet al., 2023 ] Shunyu Yao, Dian Yu, Jeffrey Zhao, Izhak Shafran, Tom Griffiths, Yuan Cao, and Karthik Narasimhan. Tree of thoughts: Deliberate problem solving with large language models.NeurIPS, 36:11809–11822,

    2023

  31. [39]

    Rlhf-v: To- wards trustworthy mllms via behavior alignment from fine-grained correctional human feedback

    [Yuet al., 2024 ] Tianyu Yu, Yuan Yao, Haoye Zhang, Tai- wen He, Yifeng Han, Ganqu Cui, Jinyi Hu, Zhiyuan Liu, Hai-Tao Zheng, Maosong Sun, et al. Rlhf-v: To- wards trustworthy mllms via behavior alignment from fine-grained correctional human feedback. InCVPR, pages 13807–13816,

    2024

  32. [40]

    Least-to-most prompting enables complex reasoning in large language models

    [Zhouet al., 2023 ] Denny Zhou, Nathanael Sch¨arli, Le Hou, Jason Wei, Nathan Scales, Xuezhi Wang, Dale Schuur- mans, Claire Cui, Olivier Bousquet, Quoc V Le, et al. Least-to-most prompting enables complex reasoning in large language models. InICLR,

    2023

  33. [41]

    Diccr: Double-gated intervention and confounder causal reasoning for vision-language nav- igation.Neural Networks, 184:107078,

    [Zhouet al., 2025 ] Dongming Zhou, Jinsheng Deng, Zheng- bin Pang, and Wei Li. Diccr: Double-gated intervention and confounder causal reasoning for vision-language nav- igation.Neural Networks, 184:107078,

    2025

  34. [42]

    Minigpt-4: Enhancing vision-language understanding with advanced large lan- guage models.arXiv preprint arXiv:2304.10592, 2023

    [Zhuet al., 2023 ] Deyao Zhu, Jun Chen, Xiaoqian Shen, Xi- ang Li, and Mohamed Elhoseiny. Minigpt-4: Enhancing vision-language understanding with advanced large lan- guage models.arXiv preprint arXiv:2304.10592, 2023

    2023 arXiv

Pith tools

Reviewed July 8, 2026 · model on record in the stance chip above.