Search

Existing vision-language understanding benchmarks largely consist of images of objects in their usual contexts. As a consequence, recent multimodal large language models can perform well with only a shallow visual understanding by relying on background language biases. Thus, strong performance on these benchmarks does not necessarily correlate with strong visual understanding. In this paper, we release JourneyBench, a comprehensive human-annotated benchmark of generated images designed to assess the model's fine-grained multimodal reasoning abilities across five tasks: complementary multimodal chain of thought, multi-image VQA, imaginary image captioning, VQA with hallucination triggers, and fine-grained retrieval with sample-specific distractors. Unlike existing benchmarks, JourneyBench explicitly requires fine-grained multimodal reasoning in unusual imaginary scenarios where language bias and holistic image gist are insufficient. We benchmark state-of-the-art models on JourneyBench and analyze performance along a number of fine-grained dimensions. Results across all five tasks show that JourneyBench is exceptionally challenging for even the best models, indicating that models' visual reasoning abilities are not as strong as they first appear. We discuss the implications of our findings and propose avenues for further research.

Semantic segmentation of night-time images holds significant importance in computer vision, particularly for applications like night environment perception in autonomous driving systems. However, existing methods tend to parse night-time images from a day-time perspective, leaving the inherent challenges in low-light conditions (such as compromised texture and deceiving matching errors) unexplored. To address these issues, we propose a novel end-to-end optimized approach, named NightFormer, tailored for night-time semantic segmentation, avoiding the conventional practice of forcibly fitting night-time images into day-time distributions. Specifically, we design a pixel-level texture enhancement module to acquire texture-aware features hierarchically with phase enhancement and amplified attention, and an object-level reliable matching module to realize accurate association matching via reliable attention in low-light environments. Extensive experimental results on various challenging benchmarks including NightCity, BDD and Cityscapes demonstrate that our proposed method performs favorably against state-of-the-art night-time semantic segmentation methods., Comment: ECCV 2024

Point cloud few-shot semantic segmentation (PC-FSS) aims to segment targets of novel categories in a given query point cloud with only a few annotated support samples. The current top-performing prototypical learning methods employ prototypes originating from support samples to direct the classification of query points. However, the inherent fragility of point-level matching and the prevalent intra-class diversity pose great challenges to this cross-instance matching paradigm, leading to erroneous background activations or incomplete target excavation. In this work, we propose a simple yet effective framework in the spirit of Decoupled Localization and Expansion (DLE). The proposed DLE, including a structural localization module (SLM) and a self-expansion module (SEM), enjoys several merits. First, structural information is injected into the matching process through the agent-level correlation in SLM, and the confident target region can thus be precisely located. Second, more reliable intra-object similarity is harnessed in SEM to derive the complete target, and the conservative expansion strategy is introduced to reasonably constrain the expansion. Extensive experiments on two challenging benchmarks under different settings demonstrate that DLE outperforms previous state-of-the-art approaches by large margins., Comment: Accepted to ECCV 2024

Two-dimensional materials with Janus structure host novel physical properties due to their inversional symmetry breaking. However, it remains elusive to synthesize Janus monolayer crystals with tailored long-range magnetic orders. Here, we have developed a general method to fabricate uniform Janus CrTeSe monolayers by selective selenization of preformed CrTe2 monolayers with molecular beam epitaxy. The uniform Janus structure of CrTeSe with high crystal quality is confirmed by high-resolution scanning transmission electron microscopy. Spin-polarized scanning tunneling microscopy/spectroscopy measurements unveil that the Janus CrTeSe undergoes a charge density wave (CDW) transition and a robust antiferromagnetic order. The magnetic anisotropy of CrTeSe is drastically altered compared to monolayer CrTe2 by the breaking symmetries induced from the Janus structure and the CDW transition, as is substantiated with first principles calculations. Our research achieves the construction of large-area Janus structures, and artificially tailors the electronic and magnetic properties of Janus systems at the two-dimensional limit., Comment: 19 pages, 4 figures

To tackle the problem of domain-specific knowledge scarcity within large language models (LLMs), knowledge graph-retrievalaugmented method has been proven to be an effective and efficient technique for knowledge infusion. However, existing approaches face two primary challenges: knowledge mismatch between public available knowledge graphs and the specific domain of the task at hand, and poor information compliance of LLMs with knowledge graphs. In this paper, we leverage a small set of labeled samples and a large-scale corpus to efficiently construct domain-specific knowledge graphs by an LLM, addressing the issue of knowledge mismatch. Additionally, we propose a three-stage KG-LLM alignment strategyto enhance the LLM's capability to utilize information from knowledge graphs. We conduct experiments with a limited-sample setting on two biomedical question-answering datasets, and the results demonstrate that our approach outperforms existing baselines., Comment: ACL2024 Findings

We introduce a rehearsal-free federated domain incremental learning framework, RefFiL, based on a global prompt-sharing paradigm to alleviate catastrophic forgetting challenges in federated domain-incremental learning, where unseen domains are continually learned. Typical methods for mitigating forgetting, such as the use of additional datasets and the retention of private data from earlier tasks, are not viable in federated learning (FL) due to devices' limited resources. Our method, RefFiL, addresses this by learning domain-invariant knowledge and incorporating various domain-specific prompts from the domains represented by different FL participants. A key feature of RefFiL is the generation of local fine-grained prompts by our domain adaptive prompt generator, which effectively learns from local domain knowledge while maintaining distinctive boundaries on a global scale. We also introduce a domain-specific prompt contrastive learning loss that differentiates between locally generated prompts and those from other domains, enhancing RefFiL's precision and effectiveness. Compared to existing methods, RefFiL significantly alleviates catastrophic forgetting without requiring extra memory space, making it ideal for privacy-sensitive and resource-constrained devices.

Despite the widespread adoption of Vision-Language Understanding (VLU) benchmarks such as VQA v2, OKVQA, A-OKVQA, GQA, VCR, SWAG, and VisualCOMET, our analysis reveals a pervasive issue affecting their integrity: these benchmarks contain samples where answers rely on assumptions unsupported by the provided context. Training models on such data foster biased learning and hallucinations as models tend to make similar unwarranted assumptions. To address this issue, we collect contextual data for each sample whenever available and train a context selection module to facilitate evidence-based model predictions. Strong improvements across multiple benchmarks demonstrate the effectiveness of our approach. Further, we develop a general-purpose Context-AwaRe Abstention (CARA) detector to identify samples lacking sufficient context and enhance model accuracy by abstaining from responding if the required context is absent. CARA exhibits generalization to new benchmarks it wasn't trained on, underscoring its utility for future VLU benchmarks in detecting or cleaning samples with inadequate context. Finally, we curate a Context Ambiguity and Sufficiency Evaluation (CASE) set to benchmark the performance of insufficient context detectors. Overall, our work represents a significant advancement in ensuring that vision-language models generate trustworthy and evidence-based outputs in complex real-world scenarios.

With impressive achievements made, artificial intelligence is on the path forward to artificial general intelligence. Sora, developed by OpenAI, which is capable of minute-level world-simulative abilities can be considered as a milestone on this developmental path. However, despite its notable successes, Sora still encounters various obstacles that need to be resolved. In this survey, we embark from the perspective of disassembling Sora in text-to-video generation, and conducting a comprehensive review of literature, trying to answer the question, \textit{From Sora What We Can See}. Specifically, after basic preliminaries regarding the general algorithms are introduced, the literature is categorized from three mutually perpendicular dimensions: evolutionary generators, excellent pursuit, and realistic panorama. Subsequently, the widely used datasets and metrics are organized in detail. Last but more importantly, we identify several challenges and open problems in this domain and propose potential future directions for research and development., Comment: A comprehensive list of text-to-video generation studies in this survey is available at https://github.com/soraw-ai/Awesome-Text-to-Video-Generation

In this paper, we investigate the precoder design for user-centric network (UCN) massive multiple-input multiple-output (mMIMO) downlink with matrix manifold optimization. In UCN mMIMO systems, each user terminal (UT) is served by a subset of base stations (BSs) instead of all the BSs, facilitating the implementation of the system and lowering the dimension of the precoders to be designed. By proving that the precoder set satisfying the per-BS power constraints forms a Riemannian submanifold of a linear product manifold, we transform the constrained precoder design problem in Euclidean space to an unconstrained one on the Riemannian submanifold. Riemannian ingredients, including orthogonal projection, Riemannian gradient, retraction and vector transport, of the problem on the Riemannian submanifold are further derived, with which the Riemannian conjugate gradient (RCG) design method is proposed for solving the unconstrained problem. The proposed method avoids the inverses of large dimensional matrices, which is beneficial in practice. The complexity analyses show the high computational efficiency of RCG precoder design. Simulation results demonstrate the numerical superiority of the proposed precoder design and the high efficiency of the UCN mMIMO system., Comment: 13 pages, 9 figures, journal

Open-vocabulary semantic segmentation (OVS) aims to segment images of arbitrary categories specified by class labels or captions. However, most previous best-performing methods, whether pixel grouping methods or region recognition methods, suffer from false matches between image features and category labels. We attribute this to the natural gap between the textual features and visual features. In this work, we rethink how to mitigate false matches from the perspective of image-to-image matching and propose a novel relation-aware intra-modal matching (RIM) framework for OVS based on visual foundation models. RIM achieves robust region classification by firstly constructing diverse image-modal reference features and then matching them with region features based on relation-aware ranking distribution. The proposed RIM enjoys several merits. First, the intra-modal reference features are better aligned, circumventing potential ambiguities that may arise in cross-modal matching. Second, the ranking-based matching process harnesses the structure information implicit in the inter-class relationships, making it more robust than comparing individually. Extensive experiments on three benchmarks demonstrate that RIM outperforms previous state-of-the-art methods by large margins, obtaining a lead of more than 10% in mIoU on PASCAL VOC benchmark., Comment: Accepted to CVPR2024

Headline generation aims to summarize a long document with a short, catchy title that reflects the main idea. This requires accurately capturing the core document semantics, which is challenging due to the lengthy and background information-rich na ture of the texts. In this work, We propose using a unified semantic discourse structure (S3) to represent document semantics, achieved by combining document-level rhetorical structure theory (RST) trees with sentence-level abstract meaning representation (AMR) graphs to construct S3 graphs. The hierarchical composition of sentence, clause, and word intrinsically characterizes the semantic meaning of the overall document. We then develop a headline generation framework, in which the S3 graphs are encoded as contextual features. To consolidate the efficacy of S3 graphs, we further devise a hierarchical structure pruning mechanism to dynamically screen the redundant and nonessential nodes within the graph. Experimental results on two headline generation datasets demonstrate that our method outperforms existing state-of-art methods consistently. Our work can be instructive for a broad range of document modeling tasks, more than headline or summarization generation.

We determine the maximum number of a graph without containing the 2-power of a Hamilton path. Using this result, we establish a spectral condition for a graph containing the 2-power of a Hamilton path.

Augmentation is an effective alternative to utilize the small amount of labeled protein data. However, most of the existing work focuses on design-ing new architectures or pre-training tasks, and relatively little work has studied data augmentation for proteins. This paper extends data augmentation techniques previously used for images and texts to proteins and then benchmarks these techniques on a variety of protein-related tasks, providing the first comprehensive evaluation of protein augmentation. Furthermore, we propose two novel semantic-level protein augmentation methods, namely Integrated Gradients Substitution and Back Translation Substitution, which enable protein semantic-aware augmentation through saliency detection and biological knowledge. Finally, we integrate extended and proposed augmentations into an augmentation pool and propose a simple but effective framework, namely Automated Protein Augmentation (APA), which can adaptively select the most suitable augmentation combinations for different tasks. Extensive experiments have shown that APA enhances the performance of five protein related tasks by an average of 10.55% across three architectures compared to vanilla implementations without augmentation, highlighting its potential to make a great impact on the field.

To naturally allow for string compactification with duality manifested, here we investigate in the self-mirror large volume scenarios from Schoen Calabi-Yau manifold. We explicitly study the geometry of Schoen Calabi-Yau threefold and complete its triple intersection from both ambient and non-ambient spaces. Based on these, we study the large volume scenario of self-mirror Calabi-Yau compactification with Schoen type. Moreover, by studying the leading non-perturbative terms of the effective scalar potential, we find special uplift terms in order of F-term $\mathcal{O}(\frac{1}{\mathcal{V}^2})$ arising from self-mirror large volume scenario. In particular, the quotient Schoen and Schoen Calabi-Yau large volume scenarios both give rise to de Sitter vacua. In addition, we discussed on the criteria to the effective scalar potential derived from self-mirror large volume scenario according to the swampland conjecture with the constraints fulfilled., Comment: typos corrected, references added

The large volume scenario has been an important issue for flux compactifications with T-dual non-geometric fluxes. As one solution to this issue, to naturally embed duality in string compactification, we investigate in self-mirror Calabi-Yau flux compactification with large volume scenario visited. In particular, at the large volume limit, the non-perturbative terms contribute a special dominant uplift term in the order of $\mathcal{O}\left(\frac{1}{\mathcal{V}^2}\right)$, while the $\alpha'$-corrections are trivialized due to the self-mirror Calabi-Yau construction. These in total contribute to effective scalar potential in the same order as from F-term $\frac{D W. DW}{\mathcal{V}^2}$, and essentially give rise to de Sitter vacua allowed by swampland conjectures., Comment: typos corrected, discussion expanded, references added

We propose a new density estimation algorithm. Given $n$ i.i.d. samples from a distribution belonging to a class of densities on $\mathbb{R}^d$, our estimator outputs any density in the class whose ''perceptron discrepancy'' with the empirical distribution is at most $O(\sqrt{d/n})$. The perceptron discrepancy between two distributions is defined as the largest difference in mass that they place on any halfspace of $\mathbb{R}^d$. It is shown that this estimator achieves expected total variation distance to the truth that is almost minimax optimal over the class of densities with bounded Sobolev norm and Gaussian mixtures. This suggests that regularity of the prior distribution could be an explanation for the efficiency of the ubiquitous step in machine learning that replaces optimization over large function spaces with simpler parametric classes (e.g. in the discriminators of GANs). We generalize the above to show that replacing the ''perceptron discrepancy'' with the generalized energy distance of Sz\'ekeley-Rizzo further improves total variation loss. The generalized energy distance between empirical distributions is easily computable and differentiable, thus making it especially useful for fitting generative models. To the best of our knowledge, it is the first example of a distance with such properties for which there are minimax statistical guarantees., Comment: 47 pages

Recent works have shown that Large Language Models (LLMs) could empower traditional neuro-symbolic models via programming capabilities to translate language into module descriptions, thus achieving strong visual reasoning results while maintaining the model's transparency and efficiency. However, these models usually exhaustively generate the entire code snippet given each new instance of a task, which is extremely ineffective. We propose generative neuro-symbolic visual reasoning by growing and reusing modules. Specifically, our model consists of three unique stages, module initialization, module generation, and module execution. First, given a vision-language task, we adopt LLMs to examine whether we could reuse and grow over established modules to handle this new task. If not, we initialize a new module needed by the task and specify the inputs and outputs of this new module. After that, the new module is created by querying LLMs to generate corresponding code snippets that match the requirements. In order to get a better sense of the new module's ability, we treat few-shot training examples as test cases to see if our new module could pass these cases. If yes, the new module is added to the module library for future reuse. Finally, we evaluate the performance of our model on the testing set by executing the parsed programs with the newly made visual modules to get the results. We find the proposed model possesses several advantages. First, it performs competitively on standard tasks like visual question answering and referring expression comprehension; Second, the modules learned from one task can be seamlessly transferred to new tasks; Last but not least, it is able to adapt to new visual reasoning tasks by observing a few training examples and reusing modules.

While enabling large language models to implement function calling (known as APIs) can greatly enhance the performance of Large Language Models (LLMs), function calling is still a challenging task due to the complicated relations between different APIs, especially in a context-learning setting without fine-tuning. This paper introduces ``Reverse Chain'', a controllable, target-driven approach designed to empower LLMs with the capability to operate external APIs only via prompts. Recognizing that most LLMs have limited tool-use capabilities, Reverse Chain limits LLMs to executing simple tasks, e.g., API Selection and Argument Completion. Furthermore, to manage a controllable multi-function calling, Reverse Chain adopts a generic rule based on a backward reasoning process. This rule determines when to do API selection or Argument completion. To evaluate the multi-tool-use capability of LLMs, we have released a compositional multi-tool task dataset, available at \url{https://anonymous.4open.science/r/reverse-chain-8681}. Extensive numerical experiments validate the remarkable proficiency of Reverse Chain in managing multiple API calls.

With the rapid development of Large Language Models (LLMs), increasing attention has been paid to their safety concerns. Consequently, evaluating the safety of LLMs has become an essential task for facilitating the broad applications of LLMs. Nevertheless, the absence of comprehensive safety evaluation benchmarks poses a significant impediment to effectively assess and enhance the safety of LLMs. In this work, we present SafetyBench, a comprehensive benchmark for evaluating the safety of LLMs, which comprises 11,435 diverse multiple choice questions spanning across 7 distinct categories of safety concerns. Notably, SafetyBench also incorporates both Chinese and English data, facilitating the evaluation in both languages. Our extensive tests over 25 popular Chinese and English LLMs in both zero-shot and few-shot settings reveal a substantial performance advantage for GPT-4 over its counterparts, and there is still significant room for improving the safety of current LLMs. We also demonstrate that the measured safety understanding abilities in SafetyBench are correlated with safety generation abilities. Data and evaluation guidelines are available at \url{https://github.com/thu-coai/SafetyBench}{https://github.com/thu-coai/SafetyBench}. Submission entrance and leaderboard are available at \url{https://llmbench.ai/safety}{https://llmbench.ai/safety}., Comment: ACL 2024 Main Conference

Given $n$ observations from two balanced classes, consider the task of labeling an additional $m$ inputs that are known to all belong to \emph{one} of the two classes. Special cases of this problem are well-known: with complete knowledge of class distributions ($n=\infty$) the problem is solved optimally by the likelihood-ratio test; when $m=1$ it corresponds to binary classification; and when $m\approx n$ it is equivalent to two-sample testing. The intermediate settings occur in the field of likelihood-free inference, where labeled samples are obtained by running forward simulations and the unlabeled sample is collected experimentally. In recent work it was discovered that there is a fundamental trade-off between $m$ and $n$: increasing the data sample $m$ reduces the amount $n$ of training/simulation data needed. In this work we (a) introduce a generalization where unlabeled samples come from a mixture of the two classes -- a case often encountered in practice; (b) study the minimax sample complexity for non-parametric classes of densities under \textit{maximum mean discrepancy} (MMD) separation; and (c) investigate the empirical performance of kernels parameterized by neural networks on two tasks: detection of the Higgs boson and detection of planted DDPM generated images amidst CIFAR-10 images. For both problems we confirm the existence of the theoretically predicted asymmetric $m$ vs $n$ trade-off., Comment: 36 pages, 6 figures

Vision-language tasks, such as VQA, SNLI-VE, and VCR are challenging because they require the model's reasoning ability to understand the semantics of the visual world and natural language. Supervised methods working for vision-language tasks have been well-studied. However, solving these tasks in a zero-shot setting is less explored. Since Contrastive Language-Image Pre-training (CLIP) has shown remarkable zero-shot performance on image-text matching, previous works utilized its strong zero-shot ability by converting vision-language tasks into an image-text matching problem, and they mainly consider global-level matching (e.g., the whole image or sentence). However, we find visual and textual fine-grained information, e.g., keywords in the sentence and objects in the image, can be fairly informative for semantics understanding. Inspired by this, we propose a unified framework to take advantage of the fine-grained information for zero-shot vision-language learning, covering multiple tasks such as VQA, SNLI-VE, and VCR. Our experiments show that our framework outperforms former zero-shot methods on VQA and achieves substantial improvement on SNLI-VE and VCR. Furthermore, our ablation studies confirm the effectiveness and generalizability of our proposed method. Code will be available at https://github.com/ThreeSR/UniFine, Comment: 14 pages, 4 figures, ACL 2023 Findings

In recent years, Blockchain-based Online Social Media (BOSM) platforms have evolved fast due to the advancement of blockchain technology. BOSM can effectively overcome the problems of traditional social media platforms, such as a single point of trust and insufficient incentives for users, by combining a decentralized governance structure and a cryptocurrency-based incentive model, thereby attracting a large number of users and making it a crucial component of Web3. BOSM allows users to downvote low-quality content and aims to decrease the visibility of low-quality content by sorting and filtering it through downvoting. However, this feature may be maliciously exploited by some users to undermine the fairness of the incentive, reduce the quality of highly visible content, and further reduce users' enthusiasm for content creation and the attractiveness of the platform. In this paper, we study and analyze the downvoting behavior using four years of data collected from Steemit, the largest BOSM platform. We discovered that a significant number of bot accounts were actively downvoting content. In addition, we discovered that roughly 9% of the downvoting activity might be retaliatory. We did not detect any significant instances of downvoting on content for a specific topic. We believe that the findings in this paper will facilitate the future development of user behavior analysis and incentive pattern design in BOSM and Web3.

The hybrid architecture of convolution neural networks (CNN) and Transformer has been the most popular method for medical image segmentation. However, the existing networks based on the hybrid architecture suffer from two problems. First, although the CNN branch can capture image local features by using convolution operation, the vanilla convolution is unable to achieve adaptive extraction of image features. Second, although the Transformer branch can model the global information of images, the conventional self-attention only focuses on the spatial self-attention of images and ignores the channel and cross-dimensional self-attention leading to low segmentation accuracy for medical images with complex backgrounds. To solve these problems, we propose vision Transformer embrace convolutional neural networks for medical image segmentation (TEC-Net). Our network has two advantages. First, dynamic deformable convolution (DDConv) is designed in the CNN branch, which not only overcomes the difficulty of adaptive feature extraction using fixed-size convolution kernels, but also solves the defect that different inputs share the same convolution kernel parameters, effectively improving the feature expression ability of CNN branch. Second, in the Transformer branch, a (shifted)-window adaptive complementary attention module ((S)W-ACAM) and compact convolutional projection are designed to enable the network to fully learn the cross-dimensional long-range dependency of medical images with few parameters and calculations. Experimental results show that the proposed TEC-Net provides better medical image segmentation results than SOTA methods including CNN and Transformer networks. In addition, our TEC-Net requires fewer parameters and computational costs and does not rely on pre-training. The code is publicly available at https://github.com/SR0920/TEC-Net., Comment: arXiv admin note: substantial text overlap with arXiv:2306.03373

The hybrid architecture of convolutional neural networks (CNNs) and Transformer are very popular for medical image segmentation. However, it suffers from two challenges. First, although a CNNs branch can capture the local image features using vanilla convolution, it cannot achieve adaptive feature learning. Second, although a Transformer branch can capture the global features, it ignores the channel and cross-dimensional self-attention, resulting in a low segmentation accuracy on complex-content images. To address these challenges, we propose a novel hybrid architecture of convolutional neural networks hand in hand with vision Transformers (CiT-Net) for medical image segmentation. Our network has two advantages. First, we design a dynamic deformable convolution and apply it to the CNNs branch, which overcomes the weak feature extraction ability due to fixed-size convolution kernels and the stiff design of sharing kernel parameters among different inputs. Second, we design a shifted-window adaptive complementary attention module and a compact convolutional projection. We apply them to the Transformer branch to learn the cross-dimensional long-term dependency for medical images. Experimental results show that our CiT-Net provides better medical image segmentation results than popular SOTA methods. Besides, our CiT-Net requires lower parameters and less computational costs and does not rely on pre-training. The code is publicly available at https://github.com/SR0920/CiT-Net., Comment: 9 pages, 3 figures, 3 tables

It is widely reported that deep generative models can classify out-of-distribution (OOD) samples as in-distribution with high confidence. In this work, we propose a hypothesis that this phenomenon is due to the reconstruction task, which can cause the generative model to focus too much on low-level features and not enough on semantic information. To address this issue, we introduce SR-OOD, an OOD detection framework that utilizes sample repairing to encourage the generative model to learn more than just an identity map. By focusing on semantics, our framework improves OOD detection performance without external data and label information. Our experimental results demonstrate the competitiveness of our approach in detecting OOD samples.

The field of vision-and-language (VL) understanding has made unprecedented progress with end-to-end large pre-trained VL models (VLMs). However, they still fall short in zero-shot reasoning tasks that require multi-step inferencing. To achieve this goal, previous works resort to a divide-and-conquer pipeline. In this paper, we argue that previous efforts have several inherent shortcomings: 1) They rely on domain-specific sub-question decomposing models. 2) They force models to predict the final answer even if the sub-questions or sub-answers provide insufficient information. We address these limitations via IdealGPT, a framework that iteratively decomposes VL reasoning using large language models (LLMs). Specifically, IdealGPT utilizes an LLM to generate sub-questions, a VLM to provide corresponding sub-answers, and another LLM to reason to achieve the final answer. These three modules perform the divide-and-conquer procedure iteratively until the model is confident about the final answer to the main question. We evaluate IdealGPT on multiple challenging VL reasoning tasks under a zero-shot setting. In particular, our IdealGPT outperforms the best existing GPT-4-like models by an absolute 10% on VCR and 15% on SNLI-VE. Code is available at https://github.com/Hxyou/IdealGPT, Comment: 13 pages, 5 figures