PyTorchKR Papers

“Code Generation”

총 10,163편
태그·기간 등 상세 필터

카테고리 탐색

arXiv:2205.15608공식 코드 · Papers with Code

Weakly-supervised Action Transition Learning for Stochastic Human Motion Prediction

We introduce the task of action-driven stochastic human motion prediction, which aims to predict multiple plausible future motions given a sequence of action labels and a short motion history. This differs from existing works, which predict motions that either do not respect any specific action category, or follow a single action label. In particular, addressing this task requires tackling two challenges: The transitions between the different actions must be smooth; the length of the predicted motion depends on the action sequence and varies significantly across samples. As we cannot realistically expect training data to cover sufficiently diverse action transitions and motion lengths, we propose an effective training strategy consisting of combining multiple motions from different actions and introducing a weak form of supervision to encourage smooth transitions. We then design a VAE-based model conditioned on both the observed motion and the action label sequence, allowing us to generate multiple plausible future motions of varying length. We illustrate the generality of our approach by exploring its use with two different temporal encoding models, namely RNNs and Transformers. Our approach outperforms baseline models constructed by adapting state-of-the-art single action-conditioned motion generation methods and stochastic human motion prediction approaches to our new task of action-driven stochastic motion prediction. Our code is available at https://github.com/wei-mao-2019/WAT.

arXiv:2209.11178공식 코드 · Papers with Code

Poisson Flow Generative Models

We propose a new "Poisson flow" generative model (PFGM) that maps a uniform distribution on a high-dimensional hemisphere into any data distribution. We interpret the data points as electrical charges on the $z=0$ hyperplane in a space augmented with an additional dimension $z$, generating a high-dimensional electric field (the gradient of the solution to Poisson equation). We prove that if these charges flow upward along electric field lines, their initial distribution in the $z=0$ plane transforms into a distribution on the hemisphere of radius $r$ that becomes uniform in the $r \to\infty$ limit. To learn the bijective transformation, we estimate the normalized field in the augmented space. For sampling, we devise a backward ODE that is anchored by the physically meaningful additional dimension: the samples hit the unaugmented data manifold when the $z$ reaches zero. Experimentally, PFGM achieves current state-of-the-art performance among the normalizing flow models on CIFAR-10, with an Inception score of $9.68$ and a FID score of $2.35$. It also performs on par with the state-of-the-art SDE approaches while offering $10\times $ to $20 \times$ acceleration on image generation tasks. Additionally, PFGM appears more tolerant of estimation errors on a weaker network architecture and robust to the step size in the Euler method. The code is available at https://github.com/Newbeeer/poisson_flow .

arXiv:2210.01075

Decompiling x86 Deep Neural Network Executables

Due to their widespread use on heterogeneous hardware devices, deep learning (DL) models are compiled into executables by DL compilers to fully leverage low-level hardware primitives. This approach allows DL computations to be undertaken at low cost across a variety of computing platforms, including CPUs, GPUs, and various hardware accelerators. We present BTD (Bin to DNN), a decompiler for deep neural network (DNN) executables. BTD takes DNN executables and outputs full model specifications, including types of DNN operators, network topology, dimensions, and parameters that are (nearly) identical to those of the input models. BTD delivers a practical framework to process DNN executables compiled by different DL compilers and with full optimizations enabled on x86 platforms. It employs learning-based techniques to infer DNN operators, dynamic analysis to reveal network architectures, and symbolic execution to facilitate inferring dimensions and parameters of DNN operators. Our evaluation reveals that BTD enables accurate recovery of full specifications of complex DNNs with millions of parameters (e.g., ResNet). The recovered DNN specifications can be re-compiled into a new DNN executable exhibiting identical behavior to the input executable. We show that BTD can boost two representative attacks, adversarial example generation and knowledge stealing, against DNN executables. We also demonstrate cross-architecture legacy code reuse using BTD, and envision BTD being used for other critical downstream tasks like DNN security hardening and patching.

arXiv:2306.11087공식 코드 · Papers with Code

Primitive Generation and Semantic-related Alignment for Universal Zero-Shot Segmentation

We study universal zero-shot segmentation in this work to achieve panoptic, instance, and semantic segmentation for novel categories without any training samples. Such zero-shot segmentation ability relies on inter-class relationships in semantic space to transfer the visual knowledge learned from seen categories to unseen ones. Thus, it is desired to well bridge semantic-visual spaces and apply the semantic relationships to visual feature learning. We introduce a generative model to synthesize features for unseen categories, which links semantic and visual spaces as well as addresses the issue of lack of unseen training data. Furthermore, to mitigate the domain gap between semantic and visual spaces, firstly, we enhance the vanilla generator with learned primitives, each of which contains fine-grained attributes related to categories, and synthesize unseen features by selectively assembling these primitives. Secondly, we propose to disentangle the visual feature into the semantic-related part and the semantic-unrelated part that contains useful visual classification clues but is less relevant to semantic representation. The inter-class relationships of semantic-related visual features are then required to be aligned with those in semantic space, thereby transferring semantic knowledge to visual feature learning. The proposed approach achieves impressively state-of-the-art performance on zero-shot panoptic segmentation, instance segmentation, and semantic segmentation. Code is available at https://henghuiding.github.io/PADing/.

arXiv:2309.11419

KOSMOS-2.5: A Multimodal Literate Model

The automatic reading of text-intensive images represents a significant advancement toward achieving Artificial General Intelligence (AGI). In this paper we present KOSMOS-2.5, a multimodal literate model for machine reading of text-intensive images. Pre-trained on a large-scale corpus of text-intensive images, KOSMOS-2.5 excels in two distinct yet complementary transcription tasks: (1) generating spatially-aware text blocks, where each block of text is assigned spatial coordinates within the image, and (2) producing structured text output that captures both style and structure in markdown format. This unified multimodal literate capability is achieved through a shared decoder-only autoregressive Transformer architecture and task-specific prompts. Building on this foundation, we fine-tune KOSMOS-2.5 for document understanding tasks, resulting in a document understanding generalist named KOSMOS-2.5-CHAT. Additionally, a large corpus of 357.4 million document pages spanning diverse domains was curated for pre-training. We evaluate KOSMOS-2.5 on two newly proposed benchmarks, OCREval and MarkdownEval, for document-level text recognition and image-to-markdown generation, demonstrating impressive literate capabilities comparable to GPT-4o. KOSMOS-2.5-CHAT achieves performance comparable to other state-of-the-art generalists that are five times larger (1.3B vs. 7B) across nine text-rich visual question answering benchmarks. Models and code have been available at \url{https://aka.ms/kosmos25}.

arXiv:2309.15850

Reflection Invariance Learning for Few-shot Semantic Segmentation

Few-shot semantic segmentation (FSS) aims to segment objects of unseen classes in query images with only a few annotated support images. Existing FSS algorithms typically focus on mining category representations from the single-view support to match semantic objects of the single-view query. However, the limited annotated samples render the single-view matching struggle to perceive the reflection invariance of novel objects, which results in a restricted learning space for novel categories and further induces a biased segmentation with demoted parsing performance. To address this challenge, this paper proposes a fresh few-shot segmentation framework to mine the reflection invariance in a multi-view matching manner. Specifically, original and reflection support features from different perspectives with the same semantics are learnable fused to obtain the reflection invariance prototype with a stronger category representation ability. Simultaneously, aiming at providing better prior guidance, the Reflection Invariance Prior Mask Generation (RIPMG) module is proposed to integrate prior knowledge from different perspectives. Finally, segmentation predictions from varying views are complementarily merged in the Reflection Invariance Semantic Prediction (RISP) module to yield precise segmentation predictions. Extensive experiments on both PASCAL-$5^\textit{i}$ and COCO-$20^\textit{i}$ datasets demonstrate the effectiveness of our approach and show that our method could achieve state-of-the-art performance. Code is available at \url{https://anonymous.4open.science/r/RILFS-A4D1}

arXiv:2310.05060공식 코드 · Papers with Code

DeVAn: Dense Video Annotation for Video-Language Models

We present a novel human annotated dataset for evaluating the ability for visual-language models to generate both short and long descriptions for real-world video clips, termed DeVAn (Dense Video Annotation). The dataset contains 8.5K YouTube video clips of 20-60 seconds in duration and covers a wide range of topics and interests. Each video clip is independently annotated by 5 human annotators, producing both captions (1 sentence) and summaries (3-10 sentences). Given any video selected from the dataset and its corresponding ASR information, we evaluate visuallanguage models on either caption or summary generation that is grounded in both the visual and auditory content of the video. Additionally, models are also evaluated on caption- and summary-based retrieval tasks, where the summary-based retrieval task requires the identification of a target video given excerpts of a given summary. Given the novel nature of the paragraph-length video summarization task, we compared different existing evaluation metrics and their alignment with human preferences and found that model-based evaluation metrics provide more semantically-oriented and human-aligned evaluation. Finally, we benchmarked a wide range of current video-language models on DeVAn, and we aim for DeVAn to serve as a useful evaluation set in the age of large language models and complex multi-modal tasks. Code is available at https: //github.com/TK-21st/DeVAn.

arXiv:2310.06762공식 코드 · Papers with Code

TRACE: A Comprehensive Benchmark for Continual Learning in Large Language Models

Aligned large language models (LLMs) demonstrate exceptional capabilities in task-solving, following instructions, and ensuring safety. However, the continual learning aspect of these aligned LLMs has been largely overlooked. Existing continual learning benchmarks lack sufficient challenge for leading aligned LLMs, owing to both their simplicity and the models' potential exposure during instruction tuning. In this paper, we introduce TRACE, a novel benchmark designed to evaluate continual learning in LLMs. TRACE consists of 8 distinct datasets spanning challenging tasks including domain-specific tasks, multilingual capabilities, code generation, and mathematical reasoning. All datasets are standardized into a unified format, allowing for effortless automatic evaluation of LLMs. Our experiments show that after training on TRACE, aligned LLMs exhibit significant declines in both general ability and instruction-following capabilities. For example, the accuracy of llama2-chat 13B on gsm8k dataset declined precipitously from 28.8\% to 2\% after training on our datasets. This highlights the challenge of finding a suitable tradeoff between achieving performance on specific tasks while preserving the original prowess of LLMs. Empirical findings suggest that tasks inherently equipped with reasoning paths contribute significantly to preserving certain capabilities of LLMs against potential declines. Motivated by this, we introduce the Reasoning-augmented Continual Learning (RCL) approach. RCL integrates task-specific cues with meta-rationales, effectively reducing catastrophic forgetting in LLMs while expediting convergence on novel tasks.

arXiv:2310.18986공식 코드 · Papers with Code

Controllable Group Choreography using Contrastive Diffusion

Music-driven group choreography poses a considerable challenge but holds significant potential for a wide range of industrial applications. The ability to generate synchronized and visually appealing group dance motions that are aligned with music opens up opportunities in many fields such as entertainment, advertising, and virtual performances. However, most of the recent works are not able to generate high-fidelity long-term motions, or fail to enable controllable experience. In this work, we aim to address the demand for high-quality and customizable group dance generation by effectively governing the consistency and diversity of group choreographies. In particular, we utilize a diffusion-based generative approach to enable the synthesis of flexible number of dancers and long-term group dances, while ensuring coherence to the input music. Ultimately, we introduce a Group Contrastive Diffusion (GCD) strategy to enhance the connection between dancers and their group, presenting the ability to control the consistency or diversity level of the synthesized group animation via the classifier-guidance sampling technique. Through intensive experiments and evaluation, we demonstrate the effectiveness of our approach in producing visually captivating and consistent group dance motions. The experimental results show the capability of our method to achieve the desired levels of consistency and diversity, while maintaining the overall quality of the generated group choreography. The source code can be found at https://aioz-ai.github.io/GCD

arXiv:2312.09411공식 코드 · Papers with Code

OTOv3: Automatic Architecture-Agnostic Neural Network Training and Compression from Structured Pruning to Erasing Operators

Compressing a predefined deep neural network (DNN) into a compact sub-network with competitive performance is crucial in the efficient machine learning realm. This topic spans various techniques, from structured pruning to neural architecture search, encompassing both pruning and erasing operators perspectives. Despite advancements, existing methods suffers from complex, multi-stage processes that demand substantial engineering and domain knowledge, limiting their broader applications. We introduce the third-generation Only-Train-Once (OTOv3), which first automatically trains and compresses a general DNN through pruning and erasing operations, creating a compact and competitive sub-network without the need of fine-tuning. OTOv3 simplifies and automates the training and compression process, minimizes the engineering efforts required from users. It offers key technological advancements: (i) automatic search space construction for general DNNs based on dependency graph analysis; (ii) Dual Half-Space Projected Gradient (DHSPG) and its enhanced version with hierarchical search (H2SPG) to reliably solve (hierarchical) structured sparsity problems and ensure sub-network validity; and (iii) automated sub-network construction using solutions from DHSPG/H2SPG and dependency graphs. Our empirical results demonstrate the efficacy of OTOv3 across various benchmarks in structured pruning and neural architecture search. OTOv3 produces sub-networks that match or exceed the state-of-the-arts. The source code will be available at https://github.com/tianyic/only_train_once.

arXiv:2402.04236공식 코드 · Papers with Code

CogCoM: Train Large Vision-Language Models Diving into Details through Chain of Manipulations

Vision-Language Models (VLMs) have demonstrated their broad effectiveness thanks to extensive training in aligning visual instructions to responses. However, such training of conclusive alignment leads models to ignore essential visual reasoning, further resulting in failures in meticulous visual problems and unfaithful responses. Drawing inspiration from human cognition in solving visual problems (e.g., marking, zoom in), this paper introduces Chain of Manipulations, a mechanism that enables VLMs to solve problems step-by-step with evidence. After training, models can solve various visual problems by eliciting intrinsic manipulations (e.g., grounding, zoom in) with results (e.g., boxes, image) actively without involving external tools, while also allowing users to trace error causes. We study the roadmap to implement this mechanism, including (1) a flexible design of manipulations upon extensive analysis, (2) an efficient automated data generation pipeline, (3) a compatible VLM architecture capable of multi-turn multi-image, and (4) a model training process for versatile capabilities. With the design, we also manually annotate 6K high-quality samples for the challenging graphical mathematical problems. Our trained model, \textbf{CogCoM}, equipped with this mechanism with 17B parameters achieves state-of-the-art performance across 9 benchmarks from 4 categories, demonstrating the effectiveness while preserving the interpretability. Our code, model weights, and collected data are publicly available at https://github.com/THUDM/CogCoM.

arXiv:2403.09942공식 코드 · Papers with Code

Attention-Enhanced Hybrid Feature Aggregation Network for 3D Brain Tumor Segmentation

Glioblastoma is a highly aggressive and malignant brain tumor type that requires early diagnosis and prompt intervention. Due to its heterogeneity in appearance, developing automated detection approaches is challenging. To address this challenge, Artificial Intelligence (AI)-driven approaches in healthcare have generated interest in efficiently diagnosing and evaluating brain tumors. The Brain Tumor Segmentation Challenge (BraTS) is a platform for developing and assessing automated techniques for tumor analysis using high-quality, clinically acquired MRI data. In our approach, we utilized a multi-scale, attention-guided and hybrid U-Net-shaped model -- GLIMS -- to perform 3D brain tumor segmentation in three regions: Enhancing Tumor (ET), Tumor Core (TC), and Whole Tumor (WT). The multi-scale feature extraction provides better contextual feature aggregation in high resolutions and the Swin Transformer blocks improve the global feature extraction at deeper levels of the model. The segmentation mask generation in the decoder branch is guided by the attention-refined features gathered from the encoder branch to enhance the important attributes. Moreover, hierarchical supervision is used to train the model efficiently. Our model's performance on the validation set resulted in 92.19, 87.75, and 83.18 Dice Scores and 89.09, 84.67, and 82.15 Lesion-wise Dice Scores in WT, TC, and ET, respectively. The code is publicly available at https://github.com/yaziciz/GLIMS.

arXiv:2405.18776

LMO-DP: Optimizing the Randomization Mechanism for Differentially Private Fine-Tuning (Large) Language Models

Differentially Private Stochastic Gradient Descent (DP-SGD) and its variants have been proposed to ensure rigorous privacy for fine-tuning large-scale pre-trained language models. However, they rely heavily on the Gaussian mechanism, which may overly perturb the gradients and degrade the accuracy, especially in stronger privacy regimes (e.g., the privacy budget $\epsilon < 3$). To address such limitations, we propose a novel Language Model-based Optimal Differential Privacy (LMO-DP) mechanism, which takes the first step to enable the tight composition of accurately fine-tuning (large) language models with a sub-optimal DP mechanism, even in strong privacy regimes (e.g., $0.1\leq \epsilon<3$). Furthermore, we propose a novel offline optimal noise search method to efficiently derive the sub-optimal DP that significantly reduces the noise magnitude. For instance, fine-tuning RoBERTa-large (with 300M parameters) on the SST-2 dataset can achieve an accuracy of 92.20% (given $\epsilon=0.3$, $\delta=10^{-10}$) by drastically outperforming the Gaussian mechanism (e.g., $\sim 50\%$ for small $\epsilon$ and $\delta$). We also draw similar findings on the text generation tasks on GPT-2. Finally, to our best knowledge, LMO-DP is also the first solution to accurately fine-tune Llama-2 with strong differential privacy guarantees. The code will be released soon and available upon request.

arXiv:2406.18278

Generalized Deepfake Attribution

The landscape of fake media creation changed with the introduction of Generative Adversarial Networks (GAN s). Fake media creation has been on the rise with the rapid advances in generation technology, leading to new challenges in Detecting fake media. A fundamental characteristic of GAN s is their sensitivity to parameter initialization, known as seeds. Each distinct seed utilized during training leads to the creation of unique model instances, resulting in divergent image outputs despite employing the same architecture. This means that even if we have one GAN architecture, it can produce countless variations of GAN models depending on the seed used. Existing methods for attributing deepfakes work well only if they have seen the specific GAN model during training. If the GAN architectures are retrained with a different seed, these methods struggle to attribute the fakes. This seed dependency issue made it difficult to attribute deepfakes with existing methods. We proposed a generalized deepfake attribution network (GDA-N et) to attribute fake images to their respective GAN architectures, even if they are generated from a retrained version of the GAN architecture with a different seed (cross-seed) or from the fine-tuned version of the existing GAN model. Extensive experiments on cross-seed and fine-tuned data of GAN models show that our method is highly effective compared to existing methods. We have provided the source code to validate our results.

arXiv:2409.01392공식 코드 · Papers with Code

ComfyBench: Benchmarking LLM-based Agents in ComfyUI for Autonomously Designing Collaborative AI Systems

Much previous AI research has focused on developing monolithic models to maximize their intelligence, with the primary goal of enhancing performance on specific tasks. In contrast, this work attempts to study using LLM-based agents to design collaborative AI systems autonomously. To explore this problem, we first introduce ComfyBench to evaluate agents's ability to design collaborative AI systems in ComfyUI. ComfyBench is a comprehensive benchmark comprising 200 diverse tasks covering various instruction-following generation challenges, along with detailed annotations for 3,205 nodes and 20 workflows. Based on ComfyBench, we further develop ComfyAgent, a novel framework that empowers LLM-based agents to autonomously design collaborative AI systems by generating workflows. ComfyAgent is based on two core concepts. First, it represents workflows with code, which can be reversibly converted into workflows and executed as collaborative systems by the interpreter. Second, it constructs a multi-agent system that cooperates to learn from existing workflows and generate new workflows for a given task. While experimental results demonstrate that ComfyAgent achieves a comparable resolve rate to o1-preview and significantly surpasses other agents on ComfyBench, ComfyAgent has resolved only 15\% of creative tasks. LLM-based agents still have a long way to go in autonomously designing collaborative AI systems. Progress with ComfyBench is paving the way for more intelligent and autonomous collaborative AI systems.

arXiv:2410.16592

ViMGuard: A Novel Multi-Modal System for Video Misinformation Guarding

The rise of social media and short-form video (SFV) has facilitated a breeding ground for misinformation. With the emergence of large language models, significant research has gone into curbing this misinformation problem with automatic false claim detection for text. Unfortunately, the automatic detection of misinformation in SFV is a more complex problem that remains largely unstudied. While text samples are monomodal (only containing words), SFVs comprise three different modalities: words, visuals, and non-linguistic audio. In this work, we introduce Video Masked Autoencoders for Misinformation Guarding (ViMGuard), the first deep-learning architecture capable of fact-checking an SFV through analysis of all three of its constituent modalities. ViMGuard leverages a dual-component system. First, Video and Audio Masked Autoencoders analyze the visual and non-linguistic audio elements of a video to discern its intention; specifically whether it intends to make an informative claim. If it is deemed that the SFV has informative intent, it is passed through our second component: a Retrieval Augmented Generation system that validates the factual accuracy of spoken words. In evaluation, ViMGuard outperformed three cutting-edge fact-checkers, thus setting a new standard for SFV fact-checking and marking a significant stride toward trustworthy news on social platforms. To promote further testing and iteration, VimGuard was deployed into a Chrome extension and all code was open-sourced on GitHub.

arXiv:2411.12604공식 코드 · Papers with Code

SG-LRA: Self-Generating Automatic Scoliosis Cobb Angle Measurement with Low-Rank Approximation

Automatic Cobb angle measurement from X-ray images is crucial for scoliosis screening and diagnosis. However, most existing regression-based methods and segmentation-based methods struggle with inaccurate spine representations or mask connectivity/fragmentation issues. Besides, landmark-based methods suffer from insufficient training data and annotations. To address these challenges, we propose a novel framework including Self-Generation pipeline and Low-Rank Approximation representation (SG-LRA) for automatic Cobb angle measurement. Specifically, we propose a parameterized spine contour representation based on LRA, which enables eigen-spine decomposition and spine contour reconstruction. We can directly obtain spine contour with only regressed LRA coefficients, which form a more accurate spine representation than rectangular boxes. Also, we combine LRA coefficient regression with anchor box classification to solve inaccurate predictions and mask connectivity issues. Moreover, we develop a data engine with automatic annotation and automatic selection in an iterative manner, which is trained on a private Spinal2023 dataset. With our data engine, we generate the largest scoliosis X-ray dataset named Spinal-AI2024 largely without privacy leaks. Extensive experiments on public AASCE2019, private Spinal2023, and generated Spinal-AI2024 datasets demonstrate that our method achieves state-of-the-art Cobb angle measurement performance. Our code and Spinal-AI2024 dataset are available at https://github.com/Ernestchenchen/SG-LRA and https://github.com/Ernestchenchen/Spinal-AI2024, respectively.

arXiv:2412.00133공식 코드 · Papers with Code

ETAP: Event-based Tracking of Any Point

Tracking any point (TAP) recently shifted the motion estimation paradigm from focusing on individual salient points with local templates to tracking arbitrary points with global image contexts. However, while research has mostly focused on driving the accuracy of models in nominal settings, addressing scenarios with difficult lighting conditions and high-speed motions remains out of reach due to the limitations of the sensor. This work addresses this challenge with the first event camera-based TAP method. It leverages the high temporal resolution and high dynamic range of event cameras for robust high-speed tracking, and the global contexts in TAP methods to handle asynchronous and sparse event measurements. We further extend the TAP framework to handle event feature variations induced by motion -- thereby addressing an open challenge in purely event-based tracking -- with a novel feature-alignment loss which ensures the learning of motion-robust features. Our method is trained with data from a new data generation pipeline and systematically ablated across all design decisions. Our method shows strong cross-dataset generalization and performs 136% better on the average Jaccard metric than the baselines. Moreover, on an established feature tracking benchmark, it achieves a 20% improvement over the previous best event-only method and even surpasses the previous best events-and-frames method by 4.1%. Our code is available at https://github.com/tub-rip/ETAP

arXiv:2503.12974

Exploring 3D Activity Reasoning and Planning: From Implicit Human Intentions to Route-Aware Planning

3D activity reasoning and planning has attracted increasing attention in human-robot interaction and embodied AI thanks to the recent advance in multimodal learning. However, most existing studies are facing two common challenges: 1) heavy reliance on explicit instructions with little reasoning on implicit user intention; 2) negligence of inter-step route planning on robot moves. We address the above challenges by proposing 3D activity reasoning and planning, a novel 3D task that reasons the intended activities from implicit instructions and decomposes them into steps with inter-step routes and planning under the guidance of fine-grained 3D object shapes and locations from scene segmentation. We tackle the new 3D task from two perspectives. First, we construct ReasonPlan3D, a large-scale benchmark that covers diverse 3D scenes with rich implicit instructions and detailed annotations for multi-step task planning, inter-step route planning, and fine-grained segmentation. Second, we design a novel framework that introduces progressive plan generation with contextual consistency across multiple steps, as well as a scene graph that is updated dynamically for capturing critical objects and their spatial relations. Extensive experiments demonstrate the effectiveness of our benchmark and framework in reasoning activities from implicit human instructions, producing accurate stepwise task plans and seamlessly integrating route planning for multi-step moves. The dataset and code will be released.

arXiv:2503.12663공식 코드 · Papers with Code

Logic-RAG: Augmenting Large Multimodal Models with Visual-Spatial Knowledge for Road Scene Understanding

Large multimodal models (LMMs) are increasingly integrated into autonomous driving systems for user interaction. However, their limitations in fine-grained spatial reasoning pose challenges for system interpretability and user trust. We introduce Logic-RAG, a novel Retrieval-Augmented Generation (RAG) framework that improves LMMs' spatial understanding in driving scenarios. Logic-RAG constructs a dynamic knowledge base (KB) about object-object relationships in first-order logic (FOL) using a perception module, a query-to-logic embedder, and a logical inference engine. We evaluated Logic-RAG on visual-spatial queries using both synthetic and real-world driving videos. When using popular LMMs (GPT-4V, Claude 3.5) as proxies for an autonomous driving system, these models achieved only 55% accuracy on synthetic driving scenes and under 75% on real-world driving scenes. Augmenting them with Logic-RAG increased their accuracies to over 80% and 90%, respectively. An ablation study showed that even without logical inference, the fact-based context constructed by Logic-RAG alone improved accuracy by 15%. Logic-RAG is extensible: it allows seamless replacement of individual components with improved versions and enables domain experts to compose new knowledge in both FOL and natural language. In sum, Logic-RAG addresses critical spatial reasoning deficiencies in LMMs for autonomous driving applications. Code and data are available at https://github.com/Imran2205/LogicRAG.