论文检索

Miscellaneous Aspects of Machine Learning · Causality

Mathieu Chevalley, Patrick Schwab, Arash Mehrjou

Targeted and uniform interventions to a system are crucial for unveiling causal relationships. While several methods have been developed to leverage interventional data for causal structure learning, their practical application in real-world scenarios often remains challenging. Recent benchmark studies have highlighted these difficulties, even when large numbers of single-variable intervention samples are available. In this work, we demonstrate, both theoretically and empirically, that such datasets contain a wealth of causal information that can be effectively extracted under realistic assumptions about the data distribution. More specifically, we introduce a novel variant of interventional faithfulness, which relies on comparisons between the marginal distributions of each variable across observational and interventional settings, and we introduce a score on causal orders. Under this assumption, we are able to prove strong theoretical guarantees on the optimum of our score that also hold for large-scale settings. To empirically verify our theory, we introduce Intersort, an algorithm designed to infer the causal order from datasets containing large numbers of single-variable interventions by approximately optimizing our score. Intersort outperforms baselines (GIES, DCDI, PC and EASE) on almost all simulated data settings replicating common benchmarks in the field. Our proposed novel approach to modeling interventional datasets thus offers a promising avenue for advancing causal inference, highlighting significant potential for further enhancements under realistic assumptions.

Jui-Nan Yen, Si Si, Zhao Meng, Felix Yu, Venkata Sai Surya Subramanyam Duvvuri, Inderjit Dhillon, Cho-Jui Hsieh, Sanjiv Kumar

Low-rank adaption (LoRA) is a widely used parameter-efficient finetuning method for LLM that reduces memory requirements. However, current LoRA optimizers lack transformation invariance, meaning the updates depending on how the two LoRA factors are scaled or rotated. This deficiency leads to inefficient learning and sub-optimal solutions in practice. This paper introduces LoRA-RITE, a novel adaptive matrix preconditioning method for LoRA optimization, which can achieve transformation invariance and remain computationally efficient. We provide theoretical analysis to demonstrate the benefit of our method and conduct experiments on various LLM tasks with different models including Gemma 2B, 7B, and mT5-XXL. The results demonstrate consistent improvements against existing optimizers. For example, replacing Adam with LoRA-RITE during LoRA fine-tuning of Gemma-2B yielded 4.6% accuracy gain on Super-Natural Instructions and 3.5% accuracy gain across other four LLM benchmarks (HellaSwag, ArcChallenge, GSM8K, OpenBookQA).

Deep Learning · Large Language Models

Jeonghoon Shim, Gyuhyeon Seo, Cheongsu Lim, Yohan Jo

Tool-Augmented Language Models (TALMs) leverage external APIs to answer user queries across various domains. However, existing benchmark datasets for TALM research often feature simplistic dialogues that do not reflect real-world scenarios, such as the need for models to ask clarifying questions or proactively call additional APIs when essential information is missing. To address these limitations, we construct and release ToolDial, a dataset comprising 11,111 multi-turn dialogues, with an average of 8.95 turns per dialogue, based on APIs from RapidAPI. ToolDial has two key characteristics. First, the dialogues incorporate 16 user and system actions (e.g., request, clarify, fail inform) to capture the rich dynamics of real-world interactions. Second, we simulate dialogues where the system requests necessary information from the user based on API documentation and seeks additional APIs if the user fails to provide the required information. To facilitate this process, we introduce a method for generating an API graph that represents input and output compatibility between APIs. Using ToolDial, we evaluate a suite of language models on their ability to predict correct actions and extract input parameter values for API calls from the dialogue history. Modern language models achieve accuracy scores below 70\%, indicating substantial room for improvement. We provide a detailed analysis of the areas where these models fall short.

Applications · Chemistry and Drug Discovery

Alexandru Dumitrescu, Dani Korpela, Markus Heinonen, Yogesh Verma, Valerii Iakovlev, Vikas Garg, Harri Lähdesmäki

Obtaining the desired effect of drugs is highly dependent on their molecular geometries. Thus, the current prevailing paradigm focuses on 3D point-cloud atom representations, utilizing graph neural network (GNN) parametrizations, with rotational symmetries baked in via E(3) invariant layers. We prove that such models must necessarily disregard chirality, a geometric property of the molecules that cannot be superimposed on their mirror image by rotation and translation. Chirality plays a key role in determining drug safety and potency. To address this glaring issue, we introduce a novel field-based representation, proposing reference rotations that replace rotational symmetry constraints. The proposed model captures all molecular geometries including chirality, while still achieving highly competitive performance with E(3)-based methods across standard benchmarking metrics.

Applications · Computer Vision

Jiachen Qian, Hongye Yang, Shuang Wu, Jingxi Xu, Feihu Zhang

Current state-of-the-art text-to-3D generation methods struggle to produce 3D models with fine details and delicate structures due to limitations in differentiable mesh representation techniques. This limitation is particularly pronounced in anime character generation, where intricate features such as fingers, hair, and facial details are crucial for capturing the essence of the characters.In this paper, we introduce a novel, efficient, sparse differentiable mesh representation method, termed SparseCubes, alongside a sparse transformer network designed to generate high-quality 3D models. Our method significantly reduces computational requirements by over 95% and storage memory by 50%, enabling the creation of higher resolution meshes with enhanced details and delicate structures. We validate the effectiveness of our approach through its application to text-to-3D anime character generation, demonstrating its capability to accurately render subtle details and thin structures (e.g. individual fingers) in both meshes and textures.

Theory · Reinforcement Learning and Planning

Patrick Schnell, Luca Guastoni, Nils Thuerey

Temporal difference (TD) learning represents a fascinating paradox: It is the prime example of a divergent algorithm that has not vanished after its instability was proven. On the contrary, TD continues to thrive in reinforcement learning (RL), suggesting that it provides significant compensatory benefits. Empirical evidence supports this, as many RL tasks require substantial computational resources, and TD delivers a crucial speed advantage that makes these tasks solvable. However, it is limited to cases where the divergence issues are absent or negligible for unknown reasons. So far, the theoretical foundations behind the speed-up are also unclear. In our work, we address these shortcomings of TD by employing techniques for analyzing iterative schemes developed over the past century. Our analysis reveals that TD possesses a mechanism that enables efficient mapping into the smallest eigenspace—an operation previously thought to necessitate costly matrix inversion. Notably, this effect is independent of the conditioning of the problem, making it particularly well-suited for RL tasks characterized by rapidly increasing condition numbers, e.g. through delayed rewards. Our novel theoretical understanding allows us to develop a scalable algorithm that integrates TD’s speed with the reliable convergence of gradient descent (GD). We additionally validate these improvements through a rigorous mathematical proof in two dimensions, as well as experiments on problems where TD and GD falter, providing valuable insights into the future of optimization techniques in artificial intelligence

Javier Ferrando, Oscar Obeso, Senthooran Rajamanoharan, Neel Nanda

Hallucinations in large language models are a widespread problem, yet the mechanisms behind whether models will hallucinate are poorly understood, limiting our ability to solve this problem. Using sparse autoencoders as an interpretability tool, we discover that a key part of these mechanisms is entity recognition, where the model detects if an entity is one it can recall facts about. Sparse autoencoders uncover meaningful directions in the representation space, these detect whether the model recognizes an entity, e.g. detecting it doesn't know about an athlete or a movie. This shows that models can have self-knowledge: internal representations about their own capabilities. These directions are causally relevant: capable of steering the model to refuse to answer questions about known entities, or to hallucinate attributes of unknown entities when it would otherwise refuse. We demonstrate that despite the sparse autoencoders being trained on the base model, these directions have a causal effect on the chat model's refusal behavior, suggesting that chat finetuning has repurposed this existing mechanism. Furthermore, we provide an initial exploration into the mechanistic role of these directions in the model, finding that they disrupt the attention of downstream heads that typically move entity attributes to the final token.

Deep Learning · Other Representation Learning

Zhengyi Ho, Siyuan Liang, Sen Zhang, Yibing Zhan, Dacheng Tao

Hallucinations in Large Language Models (LLMs) remain a major obstacle, particularly in high-stakes applications where factual accuracy is critical. While representation editing and reading methods have made strides in reducing hallucinations, their heavy reliance on specialised tools and training on in-domain samples, makes them difficult to scale and prone to overfitting. This limits their accuracy gains and generalizability to diverse datasets. This paper presents a lightweight method, Norm Voting (NoVo), which harnesses the untapped potential of attention head norms to dramatically enhance factual accuracy in zero-shot multiple-choice questions (MCQs). NoVo begins by automatically selecting truth-correlated head norms with an efficient, inference-only algorithm using only 30 random samples, allowing NoVo to effortlessly scale to diverse datasets. Afterwards, selected head norms are employed in a simple voting algorithm, which yields significant gains in prediction accuracy. On TruthfulQA MC1, NoVo surpasses the current state-of-the-art and all previous methods by an astounding margin---at least 19 accuracy points. NoVo demonstrates exceptional generalization to 20 diverse datasets, with significant gains in over 90\% of them, far exceeding all current representation editing and reading methods. NoVo also reveals promising gains to finetuning strategies and building textual adversarial defence. NoVo's effectiveness with head norms opens new frontiers in LLM interpretability, robustness and reliability. Our code is available at: https://github.com/hozhengyi/novo

Probabilistic Methods · Bayesian Models and Methods

Yogesh Verma, Ayush Bharti, Vikas Garg

Simulation-based inference (SBI) methods typically require fully observed data to infer parameters of models with intractable likelihood functions. However, datasets often contain missing values due to incomplete observations, data corruptions (common in astrophysics), or instrument limitations (e.g., in high-energy physics applications). In such scenarios, missing data must be imputed before applying any SBI method. We formalize the problem of missing data in SBI and demonstrate that naive imputation methods can introduce bias in the estimation of SBI posterior. We also introduce a novel amortized method that addresses this issue by jointly learning the imputation model and the inference network within a neural posterior estimation (NPE) framework. Extensive empirical results on SBI benchmarks show that our approach provides robust inference outcomes compared to standard baselines for varying levels of missing data. Moreover, we demonstrate the merits of our imputation model on two real-world bioactivity datasets (Adrenergic and Kinase assays). Code is available at https://github.com/Aalto-QuML/RISE.

Applications · Neuroscience, Cognitive Science

Jiquan Wang, Sha Zhao, Zhiling Luo, Yangxuan Zhou, Haiteng Jiang, Shijian Li, Tao Li, Gang Pan

Electroencephalography (EEG) is a non-invasive technique to measure and record brain electrical activity, widely used in various BCI and healthcare applications. Early EEG decoding methods rely on supervised learning, limited by specific tasks and datasets, hindering model performance and generalizability. With the success of large language models, there is a growing body of studies focusing on EEG foundation models. However, these studies still leave challenges: Firstly, most of existing EEG foundation models employ full EEG modeling strategy. It models the spatial and temporal dependencies between all EEG patches together, but ignores that the spatial and temporal dependencies are heterogeneous due to the unique structural characteristics of EEG signals. Secondly, existing EEG foundation models have limited generalizability on a wide range of downstream BCI tasks due to varying formats of EEG data, making it challenging to adapt to. To address these challenges, we propose a novel foundation model called CBraMod. Specifically, we devise a criss-cross transformer as the backbone to thoroughly leverage the structural characteristics of EEG signals, which can model spatial and temporal dependencies separately through two parallel attention mechanisms. And we utilize an asymmetric conditional positional encoding scheme which can encode positional information of EEG patches and be easily adapted to the EEG with diverse formats. CBraMod is pre-trained on a very large corpus of EEG through patch-based masked EEG reconstruction. We evaluate CBraMod on up to 10 downstream BCI tasks (12 public datasets). CBraMod achieves the state-of-the-art performance across the wide range of tasks, proving its strong capability and generalizability. The source code is publicly available at https://github.com/wjq-learning/CBraMod.

Deep Learning · Robustness

Priyank Pathak, Shyam Marjit, Shruti Vyas, Yogesh S Rawat

Visual-language foundation Models (FMs) exhibit remarkable zero-shot generalization across diverse tasks, largely attributed to extensive pre-training on largescale datasets. However, their robustness on low-resolution/pixelated (LR) images, a common challenge in real-world scenarios, remains underexplored. We introduce LR0.FM, a comprehensive benchmark evaluating the impact of low resolution on the zero-shot classification performance of 10 FM(s) across 66 backbones and 15 datasets. We propose a novel metric, Weighted Aggregated Robustness, to address the limitations of existing metrics and better evaluate model performance across resolutions and datasets. Our key findings show that: (i) model size positively correlates with robustness to resolution degradation, (ii) pre-training dataset quality is more important than its size, and (iii) fine-tuned and higher resolution models are less robust against LR. Our analysis further reveals that the model makes semantically reasonable predictions at LR, and the lack of fine-grained details in input adversely impacts the model’s initial layers more than the deeper layers. We use these insights and introduce a simple strategy, LR-TK0, to enhance the robustness of models without compromising their pre-trained weights. We demonstrate the effectiveness of LR-TK0 for robustness against low-resolution across several datasets and its generalization capability across backbones and other approaches. Code is available at this : https://github.com/shyammarjit/LR0.FM

Applications · Everything Else

Kush Jain, Gabriel Synnaeve, Baptiste Roziere

Code generation models can help improve many common software tasks ranging from code completion to defect prediction. Most of the existing benchmarks for code generation LLMs focus on code authoring or code completion. Surprisingly, there has been far less effort dedicated to benchmarking software testing, despite the strong correlation between well-tested software and effective bug detection. To address this gap, we create and release TestGenEval, a large-scale benchmark to measure test generation performance. Based on SWEBench, TestGenEval comprises 68,647 tests from 1,210 code and test file pairs across 11 well-maintained Python repositories. It covers initial tests authoring, test suite completion, and code coverage improvements. Test authoring simulates the process of a developer writing a test suite from scratch, while test completion mimics the scenario where a developer aims to improve the coverage of an existing test suite. We evaluate several popular models, with sizes ranging from 7B to 405B parameters. Our detailed analysis highlights TestGenEval's contribution to a comprehensive evaluation of test generation performance. In particular, models struggle to generate high-coverage test suites, with the best model, GPT-4o, achieving an average coverage of only 35.2\%. This is primarily due to models struggling to reason about execution, and their frequent assertion errors when addressing complex code paths.

Deep Learning · Graph Neural Networks

Ed Davis, Ian Gallagher, Daniel Lawson, Patrick Rubin-Delanchy

Dynamic graphs provide a flexible data abstraction for modelling many sorts of real-world systems, such as transport, trade, and social networks. Graph neural networks (GNNs) are powerful tools allowing for different kinds of prediction and inference on these systems, but getting a handle on uncertainty, especially in dynamic settings, is a challenging problem.In this work we propose to use a dynamic graph representation known in the tensor literature as the unfolding, to achieve valid prediction sets via conformal prediction. This representation, a simple graph, can be input to any standard GNN and does not require any modification to existing GNN architectures or conformal prediction routines. One of our key contributions is a careful mathematical consideration of the different inference scenarios which can arise in a dynamic graph modelling context. For a range of practically relevant cases, we obtain valid prediction sets with almost no assumptions, even dispensing with exchangeability. In a more challenging scenario, which we call the semi-inductive regime, we achieve valid prediction under stronger assumptions, akin to stationarity. We provide real data examples demonstrating validity, showing improved accuracy over baselines, and sign-posting different failure modes which can occur when those assumptions are violated.

Social Aspects · Privacy-preserving Statistics and Machine Learning

Jun Luo, Chen Chen, Shandong Wu

Federated prompt learning benefits federated learning with CLIP-like Vision-Language Model's (VLM's) robust representation learning ability through prompt learning. However, current federated prompt learning methods are habitually restricted to the traditional FL paradigm, where the participating clients are generally only allowed to download a single globally aggregated model from the server. While justifiable for training full-sized models under federated settings, in this work, we argue that this paradigm is ill-suited for lightweight prompts. By facilitating the clients to download multiple pre-aggregated prompts as fixed non-local experts, we propose Personalized Federated Mixture of Adaptive Prompts (pFedMoAP), a novel FL framework that personalizes the prompt learning process through the lens of Mixture of Experts (MoE). pFedMoAP implements a local attention-based gating network that learns to generate enhanced text features for better alignment with local image data, benefiting from both local and downloaded non-local adaptive prompt experts. Extensive experiments on 9 datasets under various federated settings demonstrate the efficacy of the proposed pFedMoAP algorithm. The code is available at https://github.com/ljaiverson/pFedMoAP.

Applications · Computer Vision

Xin Gu, Yaojie Shen, Chenxi Luo, Tiejian Luo, Yan Huang, YUEWEI LIN, Heng Fan, Libo Zhang

Transformer has attracted increasing interest in spatio-temporal video grounding, or STVG, owing to its end-to-end pipeline and promising result. Existing Transformer-based STVG approaches often leverage a set of object queries, which are initialized simply using zeros and then gradually learn target position information via iterative interactions with multimodal features, for spatial and temporal localization. Despite simplicity, these zero object queries, due to lacking target-specific cues, are hard to learn discriminative target information from interactions with multimodal features in complicated scenarios (e.g., with distractors or occlusion), resulting in degradation. Addressing this, we introduce a novel $\textbf{T}$arget-$\textbf{A}$ware Transformer for $\textbf{STVG}$ ($\textbf{TA-STVG}$), which seeks to adaptively generate object queries via exploring target-specific cues from the given video-text pair, for improving STVG. The key lies in two simple yet effective modules, comprising text-guided temporal sampling (TTS) and attribute-aware spatial activation (ASA), working in a cascade. The former focuses on selecting target-relevant temporal cues from a video utilizing holistic text information, while the latter aims at further exploiting the fine-grained visual attribute information of the object from previous target-aware temporal cues, which is applied for object query initialization. Compared to existing methods leveraging zero-initialized queries, object queries in our TA-STVG, directly generated from a given video-text pair, naturally carry target-specific cues, making them adaptive and better interact with multimodal features for learning more discriminative information to improve STVG. In our experiments on three benchmarks, including HCSTVG-v1/-v2 and VidSTG, TA-STVG achieves state-of-the-art performance and significantly outperforms the baseline, validating its efficacy. Moreover, TTS and ASA are designed for general purpose. When applied to existing methods such as TubeDETR and STCAT, we show substantial performance gains, verifying its generality. Code is released at https://github.com/HengLan/TA-STVG.

Yanbiao Ma, Wei Dai, Jiayi Chen

In object detection, the number of instances is commonly used to determine whether a dataset follows a long-tailed distribution, implicitly assuming that the model will perform poorly on categories with fewer instances. This assumption has led to extensive research on category bias in datasets with imbalanced instance distributions. However, even in datasets with relatively balanced instance counts, models still exhibit bias toward certain categories, indicating that instance count alone cannot explain this phenomenon. In this work, we first introduce the concept and measurement of category informativeness. We observe a significant negative correlation between a category’s informativeness and its accuracy, suggesting that informativeness more accurately reflects the learning difficulty of a category. Based on this observation, we propose the Informativeness-Guided Angular Margin Loss (IGAM Loss), which dynamically adjusts the decision space of categories according to their informativeness, thereby mitigating category bias in long-tailed datasets. IGAM Loss not only achieves superior performance on long-tailed benchmark datasets such as LVIS v1.0 and COCO-LT but also demonstrates significant improvements for underrepresented categories in non-long-tailed datasets like Pascal VOC. Extensive experiments confirm the potential of category informativeness as a tool and the generalizability of our proposed method.

Applications · Neuroscience, Cognitive Science

Yangxuan Zhou, Sha Zhao, Jiquan Wang, Haiteng Jiang, Shijian Li, Tao Li, Gang Pan

Electroencephalography (EEG) is a non-invasive brain-computer interface technology used for recording brain electrical activity. It plays an important role in human life and has been widely uesd in real life, including sleep staging, emotion recognition, and motor imagery. However, existing EEG-related models cannot be well applied in practice, especially in clinical settings, where new patients with individual discrepancies appear every day. Such EEG-based model trained on fixed datasets cannot generalize well to the continual flow of numerous unseen subjects in real-world scenarios. This limitation can be addressed through continual learning (CL), wherein the CL model can continuously learn and advance over time. Inspired by CL, we introduce a novel Unsupervised Individual Continual Learning paradigm for handling this issue in practice. We propose the BrainUICL framework, which enables the EEG-based model to continuously adapt to the incoming new subjects. Simultaneously, BrainUICL helps the model absorb new knowledge during each adaptation, thereby advancing its generalization ability for all unseen subjects. The effectiveness of the proposed BrainUICL has been evaluated on three different mainstream EEG tasks. The BrainUICL can effectively balance both the plasticity and stability during CL, achieving better plasticity on new individuals and better stability across all the unseen individuals, which holds significance in a practical setting.

Deep Learning · Large Language Models

Eunseop Yoon, Hee Suk Yoon, Mark Hasegawa-Johnson, Chang Yoo

In the broader context of deep learning, Multimodal Large Language Models have achieved significant breakthroughs by leveraging powerful Large Language Models as a backbone to align different modalities into the language space. A prime exemplification is the development of Video Large Language Models (Video-LLMs). While numerous advancements have been proposed to enhance the video understanding capabilities of these models, they are predominantly trained on questions generated directly from video content. However, in real-world scenarios, users often pose questions that extend beyond the informational scope of the video, highlighting the need for Video-LLMs to assess the relevance of the question. We demonstrate that even the best-performing Video-LLMs fail to reject unfit questions-not necessarily due to a lack of video understanding, but because they have not been trained to identify and refuse such questions. To address this limitation, we propose alignment for answerability, a framework that equips Video-LLMs with the ability to evaluate the relevance of a question based on the input video and appropriately decline to answer when the question exceeds the scope of the video, as well as an evaluation framework with a comprehensive set of metrics designed to measure model behavior before and after alignment. Furthermore, we present a pipeline for creating a dataset specifically tailored for alignment for answerability, leveraging existing video-description paired datasets.

Deep Learning · Large Language Models

Fangyu Lei, Jixuan Chen, Yuxiao Ye, Ruisheng Cao, Dongchan Shin, Hongjin SU, Zhaoqing Suo, Hongcheng Gao, Wenjing Hu, Pengcheng Yin 等

Real-world enterprise text-to-SQL workflows often involve complex cloud or local data across various database systems, multiple SQL queries in various dialects, and diverse operations from data transformation to analytics.We introduce Spider 2.0, an evaluation framework comprising $632$ real-world text-to-SQL workflow problems derived from enterprise-level database use cases. The databases in Spider 2.0 are sourced from real data applications, often containing over 1,000 columns and stored in local or cloud database systems such as BigQuery and Snowflake.We show that solving problems in Spider 2.0 frequently requires understanding and searching through database metadata, dialect documentation, and even project-level codebases. This challenge calls for models to interact with complex SQL workflow environments, process extremely long contexts, perform intricate reasoning, and generate multiple SQL queries with diverse operations, often exceeding $100$ lines, which goes far beyond traditional text-to-SQL challenges.Our evaluations indicate that based on o1-preview, our code agent framework successfully solves only 21.3\% of the tasks, compared with 91.2\% on Spider 1.0 and 73.0\% on BIRD.Our results on Spider 2.0 show that while language models have demonstrated remarkable performance in code generation --- especially in prior text-to-SQL benchmarks --- they require significant improvement in order to achieve adequate performance for real-world enterprise usage.Progress on Spider 2.0 represents crucial steps towards developing intelligent, autonomous, code agents for real-world enterprise settings.Our code, baseline models, and data are available at [spider2-sql.github.io](spider2-sql.github.io) .

Applications · Everything Else

Haotian Xia, Zhengbang Yang, Junbo Zou, Rhys Tracy, Yuqing Wang, Chi Lu, Christopher Lai, Yanjun He, Xun Shao, Zhuoqing Xie 等

Multimodal Large Language Models (MLLMs) are advancing the ability to reason about complex sports scenarios by integrating textual and visual information. To comprehensively evaluate their capabilities, we introduce SPORTU, a benchmark designed to assess MLLMs across multi-level sports reasoning tasks. SPORTU comprises two key components: SPORTU-text, featuring 900 multiple-choice questions with human-annotated explanations for rule comprehension and strategy understanding. This component focuses on testing models' ability to reason about sports solely through question-answering (QA), without requiring visual inputs; SPORTU-video, consisting of 1,701 slow-motion video clips across 7 different sports and 12,048 QA pairs, designed to assess multi-level reasoning, from simple sports recognition to complex tasks like foul detection and rule application. We evaluated four prevalent LLMs mainly utilizing few-shot learning paradigms supplemented by chain-of-thought (CoT) prompting on the SPORTU-text part. GPT-4o achieves the highest accuracy of 71\%, but still falls short of human-level performance, highlighting room for improvement in rule comprehension and reasoning. The evaluation for the SPORTU-video part includes 6 proprietary and 8 open-source MLLMs. Experiments show that models fall short on hard tasks that require deep reasoning and rule-based understanding. GPT-4o performs the best with only 57.8\% accuracy on the hard task, showing large room for improvement. We hope that SPORTU will serve as a critical step toward evaluating models' capabilities in sports understanding and reasoning. The dataset is available at [https://github.com/chili-lab/SPORTU](https://github.com/chili-lab/SPORTU).