PyTorchKR Papers

최신 논문

총 9,187편
태그·기간 등 상세 필터

카테고리 탐색

로그인 후 이 태그 구독

arXiv:2504.01248

Automated Factual Benchmarking for In-Car Conversational Systems using Large Language Models

In-car conversational systems bring the promise to improve the in-vehicle user experience. Modern conversational systems are based on Large Language Models (LLMs), which makes them prone to errors such as hallucinations, i.e., inaccurate, fictitious, and therefore factually incorrect information. In this paper, we present an LLM-based methodology for the automatic factual benchmarking of in-car conversational systems. We instantiate our methodology with five LLM-based methods, leveraging ensembling techniques and diverse personae to enhance agreement and minimize hallucinations. We use our methodology to evaluate CarExpert, an in-car retrieval-augmented conversational question answering system, with respect to the factual correctness to a vehicle's manual. We produced a novel dataset specifically created for the in-car domain, and tested our methodology against an expert evaluation. Our results show that the combination of GPT-4 with the Input Output Prompting achieves over 90 per cent factual correctness agreement rate with expert evaluations, other than being the most efficient approach yielding an average response time of 4.5s. Our findings suggest that LLM-based testing constitutes a viable approach for the validation of conversational systems regarding their factual correctness.

arXiv:2504.01786공식 코드 · Papers with Code

BlenderGym: Benchmarking Foundational Model Systems for Graphics Editing

3D graphics editing is crucial in applications like movie production and game design, yet it remains a time-consuming process that demands highly specialized domain expertise. Automating this process is challenging because graphical editing requires performing a variety of tasks, each requiring distinct skill sets. Recently, vision-language models (VLMs) have emerged as a powerful framework for automating the editing process, but their development and evaluation are bottlenecked by the lack of a comprehensive benchmark that requires human-level perception and presents real-world editing complexity. In this work, we present BlenderGym, the first comprehensive VLM system benchmark for 3D graphics editing. BlenderGym evaluates VLM systems through code-based 3D reconstruction tasks. We evaluate closed- and open-source VLM systems and observe that even the state-of-the-art VLM system struggles with tasks relatively easy for human Blender users. Enabled by BlenderGym, we study how inference scaling techniques impact VLM's performance on graphics editing tasks. Notably, our findings reveal that the verifier used to guide the scaling of generation can itself be improved through inference scaling, complementing recent insights on inference scaling of LLM generation in coding and math tasks. We further show that inference compute is not uniformly effective and can be optimized by strategically distributing it between generation and verification.

arXiv:2504.00255공식 코드 · Papers with Code

SciReplicate-Bench: Benchmarking LLMs in Agent-driven Algorithmic Reproduction from Research Papers

This study evaluates large language models (LLMs) in generating code from algorithm descriptions from recent NLP papers. The task requires two key competencies: (1) algorithm comprehension: synthesizing information from papers and academic literature to understand implementation logic, and (2) coding expertise: identifying dependencies and correctly implementing necessary APIs. To facilitate rigorous evaluation, we introduce SciReplicate-Bench, a benchmark of 100 tasks from 36 NLP papers published in 2024, featuring detailed annotations and comprehensive test cases. Building on SciReplicate-Bench, we propose Sci-Reproducer, a multi-agent framework consisting of a Paper Agent that interprets algorithmic concepts from literature and a Code Agent that retrieves dependencies from repositories and implement solutions. To assess algorithm understanding, we introduce reasoning graph accuracy, which quantifies similarity between generated and reference reasoning graphs derived from code comments and structure. For evaluating implementation quality, we employ execution accuracy, CodeBLEU, and repository dependency/API recall metrics. In our experiments, we evaluate various powerful Non-Reasoning LLMs and Reasoning LLMs as foundational models. The best-performing LLM using Sci-Reproducer achieves only 39% execution accuracy, highlighting the benchmark's difficulty.Our analysis identifies missing or inconsistent algorithm descriptions as key barriers to successful reproduction. We will open-source our benchmark, and code at https://github.com/xyzCS/SciReplicate-Bench.

arXiv:2504.00025

Generalization Bias in Large Language Model Summarization of Scientific Research

Artificial intelligence chatbots driven by large language models (LLMs) have the potential to increase public science literacy and support scientific research, as they can quickly summarize complex scientific information in accessible terms. However, when summarizing scientific texts, LLMs may omit details that limit the scope of research conclusions, leading to generalizations of results broader than warranted by the original study. We tested 10 prominent LLMs, including ChatGPT-4o, ChatGPT-4.5, DeepSeek, LLaMA 3.3 70B, and Claude 3.7 Sonnet, comparing 4900 LLM-generated summaries to their original scientific texts. Even when explicitly prompted for accuracy, most LLMs produced broader generalizations of scientific results than those in the original texts, with DeepSeek, ChatGPT-4o, and LLaMA 3.3 70B overgeneralizing in 26 to 73% of cases. In a direct comparison of LLM-generated and human-authored science summaries, LLM summaries were nearly five times more likely to contain broad generalizations (OR = 4.85, 95% CI [3.06, 7.70]). Notably, newer models tended to perform worse in generalization accuracy than earlier ones. Our results indicate a strong bias in many widely used LLMs towards overgeneralizing scientific conclusions, posing a significant risk of large-scale misinterpretations of research findings. We highlight potential mitigation strategies, including lowering LLM temperature settings and benchmarking LLMs for generalization accuracy.

arXiv:2504.01908공식 코드 · Papers with Code

Benchmarking Synthetic Tabular Data: A Multi-Dimensional Evaluation Framework

Evaluating the quality of synthetic data remains a key challenge for ensuring privacy and utility in data-driven research. In this work, we present an evaluation framework that quantifies how well synthetic data replicates original distributional properties while ensuring privacy. The proposed approach employs a holdout-based benchmarking strategy that facilitates quantitative assessment through low- and high-dimensional distribution comparisons, embedding-based similarity measures, and nearest-neighbor distance metrics. The framework supports various data types and structures, including sequential and contextual information, and enables interpretable quality diagnostics through a set of standardized metrics. These contributions aim to support reproducibility and methodological consistency in benchmarking of synthetic data generation techniques. The code of the framework is available at https://github.com/mostly-ai/mostlyai-qa.

arXiv:2504.01905

Accelerating IoV Intrusion Detection: Benchmarking GPU-Accelerated vs CPU-Based ML Libraries

The Internet of Vehicles (IoV) may face challenging cybersecurity attacks that may require sophisticated intrusion detection systems, necessitating a rapid development and response system. This research investigates the performance advantages of GPU-accelerated libraries (cuML) compared to traditional CPU-based implementations (scikit-learn), focusing on the speed and efficiency required for machine learning models used in IoV threat detection environments. The comprehensive evaluations conducted employ four machine learning approaches (Random Forest, KNN, Logistic Regression, XGBoost) across three distinct IoV security datasets (OTIDS, GIDS, CICIoV2024). Our findings demonstrate that GPU-accelerated implementations dramatically improved computational efficiency, with training times reduced by a factor of up to 159 and prediction speeds accelerated by up to 95 times compared to traditional CPU processing, all while preserving detection accuracy. This remarkable performance breakthrough empowers researchers and security specialists to harness GPU acceleration for creating faster, more effective threat detection systems that meet the urgent real-time security demands of today's connected vehicle networks.

arXiv:2504.00694

On Benchmarking Code LLMs for Android Malware Analysis

Large Language Models (LLMs) have demonstrated strong capabilities in various code intelligence tasks. However, their effectiveness for Android malware analysis remains underexplored. Decompiled Android malware code presents unique challenges for analysis, due to the malicious logic being buried within a large number of functions and the frequent lack of meaningful function names. This paper presents CAMA, a benchmarking framework designed to systematically evaluate the effectiveness of Code LLMs in Android malware analysis. CAMA specifies structured model outputs to support key malware analysis tasks, including malicious function identification and malware purpose summarization. Built on these, it integrates three domain-specific evaluation metrics (consistency, fidelity, and semantic relevance), enabling rigorous stability and effectiveness assessment and cross-model comparison. We construct a benchmark dataset of 118 Android malware samples from 13 families collected in recent years, encompassing over 7.5 million distinct functions, and use CAMA to evaluate four popular open-source Code LLMs. Our experiments provide insights into how Code LLMs interpret decompiled code and quantify the sensitivity to function renaming, highlighting both their potential and current limitations in malware analysis.

arXiv:2503.22740

CSPO: Cross-Market Synergistic Stock Price Movement Forecasting with Pseudo-volatility Optimization

The stock market, as a cornerstone of the financial markets, places forecasting stock price movements at the forefront of challenges in quantitative finance. Emerging learning-based approaches have made significant progress in capturing the intricate and ever-evolving data patterns of modern markets. With the rapid expansion of the stock market, it presents two characteristics, i.e., stock exogeneity and volatility heterogeneity, that heighten the complexity of price forecasting. Specifically, while stock exogeneity reflects the influence of external market factors on price movements, volatility heterogeneity showcases the varying difficulty in movement forecasting against price fluctuations. In this work, we introduce the framework of Cross-market Synergy with Pseudo-volatility Optimization (CSPO). Specifically, CSPO implements an effective deep neural architecture to leverage external futures knowledge. This enriches stock embeddings with cross-market insights and thus enhances the CSPO's predictive capability. Furthermore, CSPO incorporates pseudo-volatility to model stock-specific forecasting confidence, enabling a dynamic adaptation of its optimization process to improve accuracy and robustness. Our extensive experiments, encompassing industrial evaluation and public benchmarking, highlight CSPO's superior performance over existing methods and effectiveness of all proposed modules contained therein.

arXiv:2503.21966

Benchmarking Deep Learning-Based Methods for Irradiance Nowcasting with Sky Images

To address the high levels of uncertainty associated with photovoltaic energy, an increasing number of studies focusing on short-term solar forecasting have been published. Most of these studies use deep learning-based models to directly forecast a solar irradiance or photovoltaic power value given an input of sky image sequences. Recently, however, advances in generative modeling have led to approaches that divide the forecasting problem into two sub-problems: 1) future event prediction, i.e. generating future sky images; and 2) solar irradiance or photovoltaic power nowcasting, i.e. predicting the concurrent value from a single image. One such approach is the SkyGPT model, where they show that the potential for improvement is much larger for the nowcasting model than for the generative model. Thus, in this paper, we focus on the solar irradiance nowcasting problem and conduct an extensive benchmark of deep learning architectures across the widely-used Folsom, SIRTA and NREL datasets. Moreover, we perform ablation experiments on different training configurations and data processing techniques, including the choice of the target variable used for training and adjustments of the timestamp alignment between images and irradiance measurements. In particular, we draw attention to a potential error associated with the sky image timestamps in the Folsom dataset and a possible fix is discussed. All our results are reported in terms of both the root mean squared error and the mean absolute error and, by leveraging the three datasets, we demonstrate that our findings are consistent across different solar stations.

arXiv:2503.23060공식 코드 · Papers with Code

Unsupervised Anomaly Detection in Multivariate Time Series across Heterogeneous Domains

The widespread adoption of digital services, along with the scale and complexity at which they operate, has made incidents in IT operations increasingly more likely, diverse, and impactful. This has led to the rapid development of a central aspect of "Artificial Intelligence for IT Operations" (AIOps), focusing on detecting anomalies in vast amounts of multivariate time series data generated by service entities. In this paper, we begin by introducing a unifying framework for benchmarking unsupervised anomaly detection (AD) methods, and highlight the problem of shifts in normal behaviors that can occur in practical AIOps scenarios. To tackle anomaly detection under domain shift, we then cast the problem in the framework of domain generalization and propose a novel approach, Domain-Invariant VAE for Anomaly Detection (DIVAD), to learn domain-invariant representations for unsupervised anomaly detection. Our evaluation results using the Exathlon benchmark show that the two main DIVAD variants significantly outperform the best unsupervised AD method in maximum performance, with 20% and 15% improvements in maximum peak F1-scores, respectively. Evaluation using the Application Server Dataset further demonstrates the broader applicability of our domain generalization methods.

arXiv:2503.23621

Simple Feedfoward Neural Networks are Almost All You Need for Time Series Forecasting

Time series data are everywhere -- from finance to healthcare -- and each domain brings its own unique complexities and structures. While advanced models like Transformers and graph neural networks (GNNs) have gained popularity in time series forecasting, largely due to their success in tasks like language modeling, their added complexity is not always necessary. In our work, we show that simple feedforward neural networks (SFNNs) can achieve performance on par with, or even exceeding, these state-of-the-art models, while being simpler, smaller, faster, and more robust. Our analysis indicates that, in many cases, univariate SFNNs are sufficient, implying that modeling interactions between multiple series may offer only marginal benefits. Even when inter-series relationships are strong, a basic multivariate SFNN still delivers competitive results. We also examine some key design choices and offer guidelines on making informed decisions. Additionally, we critique existing benchmarking practices and propose an improved evaluation protocol. Although SFNNs may not be optimal for every situation (hence the ``almost'' in our title) they serve as a strong baseline that future time series forecasting methods should always be compared against.

arXiv:2503.23644

Uni-Render: A Unified Accelerator for Real-Time Rendering Across Diverse Neural Renderers

Recent advancements in neural rendering technologies and their supporting devices have paved the way for immersive 3D experiences, significantly transforming human interaction with intelligent devices across diverse applications. However, achieving the desired real-time rendering speeds for immersive interactions is still hindered by (1) the lack of a universal algorithmic solution for different application scenarios and (2) the dedication of existing devices or accelerators to merely specific rendering pipelines. To overcome this challenge, we have developed a unified neural rendering accelerator that caters to a wide array of typical neural rendering pipelines, enabling real-time and on-device rendering across different applications while maintaining both efficiency and compatibility. Our accelerator design is based on the insight that, although neural rendering pipelines vary and their algorithm designs are continually evolving, they typically share common operators, predominantly executing similar workloads. Building on this insight, we propose a reconfigurable hardware architecture that can dynamically adjust dataflow to align with specific rendering metric requirements for diverse applications, effectively supporting both typical and the latest hybrid rendering pipelines. Benchmarking experiments and ablation studies on both synthetic and real-world scenes demonstrate the effectiveness of the proposed accelerator. The proposed unified accelerator stands out as the first solution capable of achieving real-time neural rendering across varied representative pipelines on edge devices, potentially paving the way for the next generation of neural graphics applications.

arXiv:2503.23145

CodeARC: Benchmarking Reasoning Capabilities of LLM Agents for Inductive Program Synthesis

Inductive program synthesis, or programming by example, requires synthesizing functions from input-output examples that generalize to unseen inputs. While large language model agents have shown promise in programming tasks guided by natural language, their ability to perform inductive program synthesis is underexplored. Existing evaluation protocols rely on static sets of examples and held-out tests, offering no feedback when synthesized functions are incorrect and failing to reflect real-world scenarios such as reverse engineering. We propose CodeARC, the Code Abstraction and Reasoning Challenge, a new evaluation framework where agents interact with a hidden target function by querying it with new inputs, synthesizing candidate functions, and iteratively refining their solutions using a differential testing oracle. This interactive setting encourages agents to perform function calls and self-correction based on feedback. We construct the first large-scale benchmark for general-purpose inductive program synthesis, featuring 1114 functions. Among 18 models evaluated, o3-mini performs best with a success rate of 52.7%, highlighting the difficulty of this task. Fine-tuning LLaMA-3.1-8B-Instruct on curated synthesis traces yields up to a 31% relative performance gain. CodeARC provides a more realistic and challenging testbed for evaluating LLM-based program synthesis and inductive reasoning.

arXiv:2503.23101

RL2Grid: Benchmarking Reinforcement Learning in Power Grid Operations

Reinforcement learning (RL) can transform power grid operations by providing adaptive and scalable controllers essential for grid decarbonization. However, existing methods struggle with the complex dynamics, aleatoric uncertainty, long-horizon goals, and hard physical constraints that occur in real-world systems. This paper presents RL2Grid, a benchmark designed in collaboration with power system operators to accelerate progress in grid control and foster RL maturity. Built on a power simulation framework developed by RTE France, RL2Grid standardizes tasks, state and action spaces, and reward structures within a unified interface for a systematic evaluation and comparison of RL approaches. Moreover, we integrate real control heuristics and safety constraints informed by the operators' expertise to ensure RL2Grid aligns with grid operation requirements. We benchmark popular RL baselines on the grid control tasks represented within RL2Grid, establishing reference performance metrics. Our results and discussion highlight the challenges that power grids pose for RL methods, emphasizing the need for novel algorithms capable of handling real-world physical systems.

arXiv:2503.23708

Towards Benchmarking and Assessing the Safety and Robustness of Autonomous Driving on Safety-critical Scenarios

Autonomous driving has made significant progress in both academia and industry, including performance improvements in perception task and the development of end-to-end autonomous driving systems. However, the safety and robustness assessment of autonomous driving has not received sufficient attention. Current evaluations of autonomous driving are typically conducted in natural driving scenarios. However, many accidents often occur in edge cases, also known as safety-critical scenarios. These safety-critical scenarios are difficult to collect, and there is currently no clear definition of what constitutes a safety-critical scenario. In this work, we explore the safety and robustness of autonomous driving in safety-critical scenarios. First, we provide a definition of safety-critical scenarios, including static traffic scenarios such as adversarial attack scenarios and natural distribution shifts, as well as dynamic traffic scenarios such as accident scenarios. Then, we develop an autonomous driving safety testing platform to comprehensively evaluate autonomous driving systems, encompassing not only the assessment of perception modules but also system-level evaluations. Our work systematically constructs a safety verification process for autonomous driving, providing technical support for the industry to establish standardized test framework and reduce risks in real-world road deployment.

arXiv:2503.23487

Benchmarking Systematic Relational Reasoning with Large Language and Reasoning Models

Large Language Models (LLMs) have been found to struggle with systematic reasoning. Even on tasks where they appear to perform well, their performance often depends on shortcuts, rather than on genuine reasoning abilities, leading them to collapse on out-of-distribution examples. Post-training strategies based on reinforcement learning and chain-of-thought prompting have recently been hailed as a step change. However, little is still known about the potential of the resulting ``Large Reasoning Models'' (LRMs) beyond problem solving in mathematics and programming, where finding genuine out-of-distribution problems can be difficult. In this paper, we focus on tasks that require systematic reasoning about relational compositions, especially for qualitative spatial and temporal reasoning. These tasks allow us to control the difficulty of problem instances, and measure in a precise way to what extent models can generalise. We find that that the considered LLMs and LRMs overall perform poorly overall, albeit better than random chance.

arXiv:2503.22567

Benchmarking Ultra-Low-Power $μ$NPUs

Efficient on-device neural network (NN) inference has various advantages over cloud-based processing, including predictable latency, enhanced privacy, greater reliability, and reduced operating costs for vendors. This has sparked the recent rapid development of microcontroller-scale NN accelerators, often referred to as neural processing units ($\mu$NPUs), designed specifically for ultra-low-power applications. In this paper we present the first comparative evaluation of a number of commercially-available $\mu$NPUs, as well as the first independent benchmarks for several of these platforms. We develop and open-source a model compilation framework to enable consistent benchmarking of quantized models across diverse $\mu$NPU hardware. Our benchmark targets end-to-end performance and includes model inference latency, power consumption, and memory overhead, alongside other factors. The resulting analysis uncovers both expected performance trends as well as surprising disparities between hardware specifications and actual performance, including $\mu$NPUs exhibiting unexpected scaling behaviors with increasing model complexity. Our framework provides a foundation for further evaluation of $\mu$NPU platforms alongside valuable insights for both hardware designers and software developers in this rapidly evolving space.

arXiv:2503.22516

Assessing Foundation Models for Sea Ice Type Segmentation in Sentinel-1 SAR Imagery

Accurate segmentation of sea ice types is essential for mapping and operational forecasting of sea ice conditions for safe navigation and resource extraction in ice-covered waters, as well as for understanding polar climate processes. While deep learning methods have shown promise in automating sea ice segmentation, they often rely on extensive labeled datasets which require expert knowledge and are time-consuming to create. Recently, foundation models (FMs) have shown excellent results for segmenting remote sensing images by utilizing pre-training on large datasets using self-supervised techniques. However, their effectiveness for sea ice segmentation remains unexplored, especially given sea ice's complex structures, seasonal changes, and unique spectral signatures, as well as peculiar Synthetic Aperture Radar (SAR) imagery characteristics including banding and scalloping noise, and varying ice backscatter characteristics, which are often missing in standard remote sensing pre-training datasets. In particular, SAR images over polar regions are acquired using different modes than used to capture the images at lower latitudes by the same sensors that form training datasets for FMs. This study evaluates ten remote sensing FMs for sea ice type segmentation using Sentinel-1 SAR imagery, focusing on their seasonal and spatial generalization. Among the selected models, Prithvi-600M outperforms the baseline models, while CROMA achieves a very similar performance in F1-score. Our contributions include offering a systematic methodology for selecting FMs for sea ice data analysis, a comprehensive benchmarking study on performances of FMs for sea ice segmentation with tailored performance metrics, and insights into existing gaps and future directions for improving domain-specific models in polar applications using SAR data.

arXiv:2503.22388공식 코드 · Papers with Code

Why Stop at One Error? Benchmarking LLMs as Data Science Code Debuggers for Multi-Hop and Multi-Bug Errors

LLMs are transforming software development, yet current code generation and code repair benchmarks mainly assess syntactic and functional correctness in simple, single-error cases. LLMs' capabilities to autonomously find and fix runtime logical errors in complex data science code remain largely unexplored. To address this gap, we introduce DSDBench: the Data Science Debugging Benchmark, the first benchmark for systematic evaluation of LLMs on multi-hop error tracing and multi-bug detection in data science code debugging. DSDBench adapts datasets from existing data science task benchmarks, such as DABench and MatPlotBench, featuring realistic data science debugging tasks with automatically synthesized multi-hop, multi-bug code snippets. DSDBench includes 1,117 annotated samples with 741 cause-effect error pairs and runtime error messages. Evaluations of state-of-the-art LLMs on DSDBench show significant performance gaps, highlighting challenges in debugging logical runtime errors in data science code. DSDBench offers a crucial resource to evaluate and improve LLMs' debugging and reasoning capabilities, enabling more reliable AI-assisted data science in the future. DSDBench is publicly available at github.com/KevinCL16/DSDBench.

arXiv:2503.22192

An Advanced Ensemble Deep Learning Framework for Stock Price Prediction Using VAE, Transformer, and LSTM Model

This research proposes a cutting-edge ensemble deep learning framework for stock price prediction by combining three advanced neural network architectures: The particular areas of interest for the research include but are not limited to: Variational Autoencoder (VAE), Transformer, and Long Short-Term Memory (LSTM) networks. The presented framework is aimed to substantially utilize the advantages of each model which would allow for achieving the identification of both linear and non-linear relations in stock price movements. To improve the accuracy of its predictions it uses rich set of technical indicators and it scales its predictors based on the current market situation. By trying out the framework on several stock data sets, and benchmarking the results against single models and conventional forecasting, the ensemble method exhibits consistently high accuracy and reliability. The VAE is able to learn linear representation on high-dimensional data while the Transformer outstandingly perform in recognizing long-term patterns on the stock price data. LSTM, based on its characteristics of being a model that can deal with sequences, brings additional improvements to the given framework, especially regarding temporal dynamics and fluctuations. Combined, these components provide exceptional directional performance and a very small disparity in the predicted results. The present solution has given a probable concept that can handle the inherent problem of stock price prediction with high reliability and scalability. Compared to the performance of individual proposals based on the neural network, as well as classical methods, the proposed ensemble framework demonstrates the advantages of combining different architectures. It has a very important application in algorithmic trading, risk analysis, and control and decision-making for finance professions and scholars.