香港中文大学(深圳)数据科学学院论文入选NeurIPS 2026

2026-11-20 17:02 来源:香港中文大学深圳招生办 评论
香港中文大学(深圳)
广东深圳市-合作办学-综合类
查看详情

9月24日,人工智能与机器学习领域国际顶级学术会议神经信息处理系统大会(Conference on Neural Information Processing Systems,简称 NeurIPS)公布了2026年论文录用结果。香港中文大学(深圳)数据科学学院师生共有28篇论文被录用,其中Main Track 23 篇、Evaluations & Datasets Track 4 篇、Position Paper Track 1 篇。 

学院2篇论文入选Oral,其中1篇来自Main Track,该赛道仅112篇入选Oral,占总投稿量的0.36%;另1篇来自Evaluations & Datasets Track,该赛道今年全球仅15篇论文入选Oral。在录用成果中,以学院学生为第一作者或共同第一作者的论文达18篇。这一优异成绩有力见证了学院在前沿领域的强劲创新动能与扎实育人成果。 

神经信息处理系统大会(NeurIPS)是人工智能与机器学习领域历史最悠久、影响力最大的国际顶级学术会议之一,与ICML、ICLR并称机器学习领域三大顶级会议。官方数据显示,本届NeurIPS主赛道共收到有效投稿30,709篇,最终录用7,900篇,录用率为25.7%;评估与数据集赛道收到有效投稿3,757篇,录用971篇,录用率为25.8%;立场论文赛道收到投稿956篇,录用135篇,录用率为22.9%。






论文及学生作者简介



*仅列举以学院学生为第一作者或共同第一作者论文

*论文作者部分:

蓝色字体为论文第一作者且为数据科学学院学生

橙色字体为学院指导教授


Main Track


题目:Incentivizing Medical Vision Capabilities from Large-Scale Multimodal Pre-training                                                                 Oral 

作者:Junying Chen, Zhenyang Cai, Yunjin Yang, Kunyuan Cai, Rongsheng   Wang, Xidong Wang, Geyang Yu, Xiangyi Feng, Benyou Wang  

摘要:Building capable medical vision-language models (VLMs) is challenging due to the scarcity of domain data and the difficulty of eliciting pre-trained medical visual knowledge. We present HuatuoGPT-5, a medical VLM built on a simple yet effective recipe: large-scale multimodal pre-training followed by reinforcement learning (RL). We first construct PubMedVision-Plus, the largest medical image-text dataset with 60M pairs from public PubMed literature, along with a clean 12M high-quality subset. After pre-training on this data, we find that traditional supervised fine-tuning (SFT) struggles to elicit the pre-trained knowledge. In contrast, by applying RL with verifiable and rubric-based instances synthesized directly from the pre-training data, we successfully incentivize the model to use this pre-trained medical visual knowledge. On PubMed-derived data, HuatuoGPT-5 outperforms similarly sized VLMs on multiple medical benchmarks and expert evaluations, significantly narrowing the gap between medical VLMs and leading proprietary VLMs. We hope these data, models, and findings can help accelerate open research on capable medical VLMs.


图片


陈俊颖

2023级 博士生 

计算机科学专业


题目:Sample Transform Cost-Based Training-Free Hallucination Detector for Large Language Models

作者:Zeyang Ding, Xinglin Hu,Jicong Fan

摘要:Hallucinations remain a major barrier to the trustworthy deployment of large language models (LLMs). We study hallucination detection from the perspective of prompt-conditioned response distributions. For a fixed prompt, an LLM induces a distribution over possible responses; when the model is uncertain or hallucinates, this distribution often becomes more complex and less internally consistent. However, this distribution is unknown, and its samples are variable-length token sequences rather than fixed-dimensional points, making direct complexity estimation difficult. To address this challenge, we propose a training-free detector based on sample-to-sample transform costs in representation space. Specifically, we represent each sampled response as an empirical distribution over generated-token hidden embeddings and compute pairwise Wasserstein distances between responses. The resulting Wasserstein distance matrix characterizes the cost structure of transforming one sampled response into another. From this matrix, we derive two complementary hallucination signals: \textbf{AvgWD}, which measures the average transform cost, and \textbf{EigenWD}, which captures the spectral complexity of the transform-cost structure. Importantly, we extend the proposed detector beyond white-box access by using an accessible auxiliary model to construct representation-space consistency signals for black-box LLMs, substantially broadening the practical applicability of our method. Experiments across five open-source LLMs and four benchmarks show that AvgWD and EigenWD consistently achieve competitive or superior AUROC compared with strong training-free uncertainty baselines. These results suggest that distributional complexity in token-level representation space provides an effective signal for hallucination detection.



图片


胡兴霖

2023级 本科生

数据科学与大数据技术专业


题目:Inexact Bregman Sparse Newton Method for Efficient Optimal Transport

作者:Jianting Pan, Ji'an Li, Sirong Dai,Ming Yan

摘要:Computing exact Optimal Transport (OT) distances for large-scale datasets is computationally prohibitive. While entropy-regularized alternatives offer speed, they sacrifice precision and frequently suffer from numerical instability in high-accuracy regimes. To address these limitations, we propose the Inexact Bregman Sparse Newton (IBSN) method, which efficiently solves the exact OT problems. Our approach utilizes a Bregman proximal point framework through a sequence of semi-dual subproblems. By solving these subproblems inexactly, we significantly reduce per-iteration complexity while maintaining a theoretical guarantee of convergence to the true optimal plan. To further accelerate the algorithm, we develop a sparse Newton-type solver for the subproblem and employ a Hessian sparsification strategy that drastically lowers memory and time costs without sacrificing accuracy. We provide rigorous theoretical guarantees for the global convergence of the algorithm. Extensive experiments demonstrate that IBSN consistently outperforms state-of-the-art methods in both computational speed and solution precision.


图片

潘建廷

2022级 博士生

数据科学专业



图片


李济安

2023级 本科生

统计学专业


图片


代司戎

2022级 博士生 

数据科学专业


题目:VersaCamVLA: Camera-Configurable VLA Policies for Robotic Manipulation

作者:Boyao Han, Chen Shi, Jingjing Qian, Zhuotao Tian, Li Jiang

摘要:Vision-Language-Action (VLA) models have emerged as powerful foundations for robotic manipulation, but their reliance on fixed camera configurations during training makes them brittle to changes in camera count or pose during deployment. To overcome these limitations, we propose VersaCamVLA, a camera-configurable framework that decouples camera-set representation from action learning. VersaCamVLA learns a unified scene-token interface that maps an arbitrary, variable set of posed RGB views into fixed-size latent scene tokens. This is achieved via multi-signal target-view prediction and Wrist-Augmented Pose Sampling (WAPS), which leverages natural wrist-camera motion for free pose diversity. At deployment, a lightweight spatial encoder injects these compact scene tokens into a pretrained base VLA as a supplementary visual condition, requiring no explicit 3D sensing or novel-view rendering. Experiments on RoboTwin, LIBERO, and a real-robot platform demonstrate that VersaCamVLA consistently outperforms prior VLA methods and direct multi-view baselines, maintaining robust performance across varying camera counts and unseen camera poses.


图片


韩伯尧

2026级 硕士生 

计算机科学专业


题目:CAPER: Clause-Aligned Process Supervision for Text-to-SQL

作者:Lujie Ban, Jiasheng Shi, Jinyang Li, Xiaolin Han, Tsz Nam Chan, Chenhao Ma

摘要:Text-to-SQL systems are typically evaluated by query-level   execution correctness, but this terminal signal provides little guidance   about which intermediate SQL decision caused success or failure. Token-level   dense supervision is also ill-suited: SQL tokens do not align with complete   semantic decisions, can penalize execution-equivalent queries, and are   difficult to label reliably at scale. We therefore propose CAPER, which   automatically derives clause-level supervision via counterfactual   intervention on the SQL abstract syntax tree (AST), enabling root-cause error   localization for reward modeling; the resulting data is used to train   CAPER-9B, a lightweight Clause-PRM that provides clause-boundary feedback for   policy optimization and candidate verification. Experiments on BIRD and   Spider show that clause-aligned supervision not only improves execution   accuracy, achieving up to a 15.3% relative EX improvement over GPT-5.4, but   also strengthens failure-localization capability, reaching 84.53% accuracy and   90.60% MRR on held-out failures.    


图片


班鲁杰

2024级 博士生

计算机科学专业


题目:Towards Generalizable Data Valuation: Learning an End-to-end Deep Model to Compute Shapley Values

作者:Zhiyu Chen,Jicong Fan

摘要:Data Shapley has emerged as a premier framework for data valuation, yet its practical utility is severely hampered by exponential computational complexity and the need for costly model retraining. Unlike existing methods that compute values independently for each dataset, we propose GenSHAP, a novel graph-based paradigm that reformulates data valuation as a transferable learning task. GenSHAP learns an amortised deep model capable of predicting Data Shapley values for any unseen classification dataset in a single forward pass. Our framework utilises a relation-based dataset representation method and a hybrid GIN-Transformer architecture to capture local and global inter-sample dependencies. We further introduce GenSHAP-S, which augments these structural representations with low-budget marginal contribution probes to achieve high-fidelity valuation. Extensive experiments demonstrate that GenSHAP closely approximates high-budget permutation-Shapley reference estimates and excels in downstream tasks such as data removal and addition while reducing valuation time from hours to seconds.

图片

陈知宇

2024级 博士生 

数据科学专业


题目:Geometric Alignment without Functional Equivalence: A Layer-wise Analysis of the Speech-Text Modality Gap

作者:Ming-Hao Hsu, Xiaohai Tian, Jun Zhang, Lu Lu, Yuxuan Wang,  Zhizheng Wu  

摘要:End-to-end speech LLMs often answer the same question less accurately from speech than from text, even when the spoken content is semantically matched to the text prompt. We study this modality gap as a layer-wise inference problem rather than a static embedding mismatch. Across four open-weight speech LLMs on SpeechMMLU and VoiceBench BBH, cross-layer Centered Kernel Alignment (CKA) with speech–text token alignment shows that speech can become geometrically text-like in middle layers while still failing to form stable late-layer answer margins. The central finding is that mid-layer alignment does not imply functional equivalence. ASR-to-LLM controls show that recognition explains part of the gap in some settings but not all of it. Matched-instance margin analysis localizes the failure to weak late-layer answer separation. Attention diagnostics further show that speech evidence remains more diffuse at the decision token. Speech-faithful temporal compaction improves BBH accuracy from 59.9% to 62.1% with tempo 1.2× and to 61.9% with silence trimming, while recovering 34.3% and 38.2% of text-correct/speech-wrong failures. These conditional repairs reveal a 2.35–2.62× asymmetry between temporal compaction and quantity-only KV merging on the same gap subset, motivating post-projection temporal abstraction rather than only input-level feature matching.


题目:Is Text All You Need? Text as a Universal Information Bottleneck for Speech LLMs    

作者:Ming-Hao Hsu, Yuxuan Hu, Shujie Liu, Jinyu Li, Zhizheng Wu 

摘要:Large language models (LLMs) provide a powerful reasoning backbone for speech understanding, but integrating continuous acoustic signals into a frozen LLM remains challenging. Existing speech-to-LLM interfaces typically operate at two extremes: either enforcing near-discrete token alignment, which benefits transcription but loses paralinguistic information, or learning unconstrained continuous representations, which can drift away from the LLM’s input space and degrade autoregressive decoding. In this work, we propose Convex Gate (C-Gate), a speech–LLM bridge that constrains all speech representations to lie within the LLM’s input embedding manifold with an architectural convex-hull constraint. Concretely, each frame is represented as a convex combination of token embeddings, ensuring compatibility with the pretrained LLM while preserving continuous expressivity. Across automatic speech recognition (ASR) and emotion recognition, C-Gate achieves strong joint performance, improving LibriSpeech WER by up to 48.7% relative while matching or exceeding single-task emotion accuracy. Beyond performance, our analysis reveals a key insight: information is not carried by discrete token identities, but by time-resolved trajectories in the embedding space. Causal interventions confirm that both the trajectory structure and alignment to the pretrained embedding manifold are critical for performance. These results suggest that geometry, rather than token discreteness, is the fundamental design factor in speech–LLM interfaces, and provide a controlled regime for studying multimodal integration in frozen LLMs. We release the checkpoint, per-sample outputs, mechanism dumps, and intervention suite for replication.


图片


徐明晧

2025级 博士生 

计算机科学专业


题目:Memory Forcing: Spatio-Temporal Memory for Consistent Scene Generation

作者:Junchao Huang, Xinting Hu, Boyao Han, Shaoshuai Shi, Zhuotao   Tian, Tianyu He, Li Jiang  

摘要:Autoregressive video diffusion models have demonstrated effectiveness in interactive game generation, with Minecraft gameplay serving as a representative application. To faithfully simulate gameplay, a model must generate natural content when exploring new scenes while preserving spatial consistency when revisiting explored areas. Under limited computation budgets, it must compress and exploit historical cues within a finite context window, which exposes a trade-off: models relying solely on temporal context offer flexible exploration but suffer from poor revisit consistency, whereas adding spatial memory strengthens consistency but may degrade new scene generation quality when the model over-relies on sparse or unreliable spatial context. We present Memory Forcing, a learning framework that pairs training protocols with a geometry-indexed spatial memory. Specifically, our Hybrid Training exposes the model to both exploration and revisitation regimes, guiding the model to rely on temporal memory in new scenes while effectively incorporating spatial memory upon revisits. Chained Forward Training creates larger pose variations and encourages reliance on spatial memory for maintaining consistency. For spatial memory, we integrate Point-to-Frame Retrieval with an incremental 3D cache generated by VGGT, enabling constant-time retrieval of relevant historical context regardless of sequence length. Extensive experiments demonstrate that Memory Forcing achieves superior performance in both long-term spatial consistency and visual quality in new scenes with real-time interaction.


图片


黄俊超

2025级 博士生 

数据科学专业


题目:Drifting Preference Optimization for One-Step Generative Models

作者:Zhou Jiang, Yandong Wen, Zhen Liu 

摘要:One-step text-to-image generators are attractive for deployment because they generate an image with a single forward pass, but preference finetuning them remains difficult: standard alignment methods often rely on policy likelihoods, denoising trajectories, differentiable reward gradients, or test-time optimization. We propose Drifting Preference Optimization (DrPO), an online preference-finetuning method for deterministic one-step generators. For each prompt, DrPO samples candidates from the current generator, ranks them with a target reward, and uses high- and low-scoring samples to synthesize a feature-space update direction. The update is a non-parametric dipole preference field plus a reference drift estimated from the frozen base generator, and is optimized through a detached feature-space regression target. The target reward is used only for ranking, so DrPO can train with large, black-box, or non-differentiable rewards while inference remains a single generator call. We evaluate DrPO on SD-Turbo and SDXL-Turbo with multiple target rewards and benchmarks, including HPSv3 and GenEval. DrPO improves alignment over reward-gradient-free one-step preference baselines and reduces HPSv3 training computation by 3.51× under the matched effective-batch setting by removing reward-model backpropagation. Initial offline experiments suggest that sample-based gradient synthesis can also be used beyond online reward ranking.


图片


姜洲

2026级 博士生 

计算机科学专业


题目:Spectral Degeneration of Softmax Attention under Isotropic Score Geometry

作者:Mengda Li, Jianfeng Yao  

摘要:We study the spectral structure of the attention matrix in Transformer models through an idealized spectral model for the pre-softmax score matrix. Motivated by the rotational isotropy of query and key representations, we replace the random token-space orientation of the score matrix by its canonical singular-value representative Σ, and assume that Σ has a logarithmic spike-bulk separation. We provide empirical diagnostics consistent with this model on pretrained vision and language transformers: token-space singular vectors exhibit near-Haar behavior, and score spectra are dominated by a small number of effective directions. We then prove that the row-wise softmax map transforms this structure into a degenerate attention matrix: dominant modes are preserved, whereas the bulk spectrum converges weakly to 𝛿_0. This provides a unified theoretical explanation for why attention layers can act as low-rank information extractors: they preserve a small number of dominant token directions while compressing the remaining bulk modes.

图片

李梦达

2025级 博士生 

数据科学专业


题目:The Reciprocity Gradient

作者:Yue Lin, Pascal Poupart, Shuhui Zhu, Dan Qiao, Wenhao Li, Yuan   Liu, Hongyuan Zha, Baoxiang Wang   

摘要:Communication is fundamental to sustaining reciprocity and cooperation in strategic interactions. We identify and formulate the influence attribution problem as the central optimization difficulty inherent in such dynamics for a learning agent: any action or signal the agent emits reshapes the reputations of many third parties along combinatorially branching paths before feeding back into its own future rewards, forcing the agent to account for all of these indirect channels at once when choosing every action. To address this, we introduce the reciprocity gradient, which explicitly backpropagates reward gradients through private estimators of opponents' policies trained from public observations. The gradient flows through the reputation chain itself analytically, rather than being estimated from sampled returns. It jointly optimizes actions and evaluative signals without intrinsic rewards or reward shaping. Empirically, the method recovers near-optimal context-sensitive policies, while sample-based baselines collapse into constant-output policies.


图片


林越

2024级 博士生 

数据科学专业


题目:Semantics-to-Contact: A Stagewise Framework for Robust   Contact-Rich Manipulation    

作者:Guanren Qiao, Ruixiang Ouyang, Sheng Xu, Yueci Deng, Yunxin Tai,   Kui Jia, Guiliang Liu   

摘要:Contact-rich manipulation requires precise control under complex contact dynamics while remaining robust to diverse visual conditions. However, real-world reinforcement learning often overfits to the visual appearance of training scenes. We propose Semantics-to-Contact (S2C), a stagewise framework for visually robust contact-rich manipulation. The insight is to decompose the problem into two stages: semantic focusing and contact refinement. In the first stage, a vision-language model localizes task-relevant regions across visually diverse scenes, providing coarse spatial guidance and reducing the burden of exploration. In the second stage, a real-world human-in-the-loop residual reinforcement learning policy learns fine-grained contact behaviors within the localized region using force feedback. To further improve visual robustness, we introduce an object-centric visual augmentation strategy that randomizes background appearance while preserving the robot and task-relevant objects. Experiments on six real-world contact-rich manipulation tasks exhibit that S2C consistently improves success rates over strong baselines and maintains reliable performance under significant visual and positional variations. These results demonstrate that stagewise semantic focusing and contact refinement provide a practical path toward visually robust real-world contact-rich manipulation.


图片


乔冠仁

2023级 博士生

计算机科学专业


题目:DriveWAM: Video Generative Priors Enable Scalable World-Action   Modeling for Autonomous Driving    

作者:Chen Shi, Jinrui Xu, Shaoshuai Shi, Kehua Sheng, Bo Zhang, Li Jiang  

摘要:Pretrained foundation models have become an important basis for end-to-end autonomous driving. In contrast to vision-language models pretrained primarily on static image-text pairs, video generative models capture temporal dynamics and motion priors that are naturally suited for driving. We present DriveWAM, a driving world-action model that adapts a pretrained video diffusion transformer into an autoregressive video-action policy. DriveWAM organizes video and action streams into a unified temporal token sequence and trains them under a joint flow-matching objective, preserving the pretrained video-generation architecture while adapting its large-scale video priors to action generation. To incorporate high-level scene understanding, we introduce scene-evolving driving guidance, where a frozen VLM produces chunk-specific semantic intent to guide video-action generation. To keep long-horizon rollout bounded, we further introduce selective KV memory, which maintains bounded modality-aware video and action memory pools through relevance-redundancy cache selection at inference time. Experiments on NAVSIM and the PhysicalAI-Autonomous-Vehicles benchmark show that DriveWAM achieves strong planning performance, and a data-scaling study from 4k to 100k driving clips further confirms the scaling potential of world-action modeling for end-to-end autonomous driving.



图片


史辰

2024级 博士生

计算机科学专业  


图片


徐锦锐

2026级 博士生

计算机科学专业 


题目:AnyEdit: A Unified Framework for Speech and Singing Voice Editing with Real-World Environmental Consistency  

作者:Yunjia Zhang, Junan Zhang, Jing Yang, Xueyao Zhang, Fan Fan,   Zhizheng Wu 

摘要:We present AnyEdit, a unified framework for speech and singing voice editing in real-world acoustic environments. Recent unified generative models already achieve strong speech and singing generation in clean conditions, but real-world vocal editing remains challenging because the model must revise content while simultaneously preserving speaker identity, prosody, and the surrounding acoustic scene. A straightforward solution is to directly model noisy edited audio end to end, yet this entangles semantic editing with environmental acoustics and can weaken the core vocal generation capability. We therefore adopt a two-stage design that preserves clean-domain editing ability while deferring acoustic rendering to a separate stage. First, an instruction-guided autoregressive editor predicts clean prosody and content-style tokens. To make this stage robust to degraded inputs, we introduce a teacher-student noise-robust prosody tokenizer that distills clean prosodic tokens from noisy recordings. Second, a flow-matching acoustic model performs in-context acoustic rendering conditioned on the source acoustic context, enabling the edited segment to inherit surrounding noise, reverberation, and background sound while maintaining speaker timbre consistently. To further stabilize the unedited regions, we introduce a unmodified-area-aware supervision mechanism. After pre-training, we also apply direct preference optimization (DPO) using real singing editing pairs mined via dynamic time warping to achieve better singing editing quality. To enable realistic evaluation, we introduce AnyEditBench, the first benchmark covering both speech and singing voice editing under diverse acoustic conditions. Experiments on Chinese and English speech and singing datasets showed that AnyEdit consistently outperforms strong baselines in content accuracy, subjective quality, and background consistency under noisy and reverberant conditions. Demo audios can be found at https://any-edit.github.io.


图片


张峻安

2024级 博士生

   计算机科学专业   


题目:MEME: Lightweight Hierarchical Mixture-of-Experts for Unified   Affective Computing    

作者:Yinan Zhang, Haoyu Zhang, Tianshu Yu   

摘要:Multimodal sentiment analysis (MSA) and emotion recognition (ER) are closely related affective understanding tasks, yet they are usually studied separately due to differences in label space, task formulation, dataset distribution, and modality dependence. In this paper, we propose MEME (Mixture-of-Experts for Multimodal sentiment analysis and Emotion recognition), a lightweight hierarchical MoE framework for unified multimodal affective computing across MSA, conversational emotion recognition, and dynamic facial expression recognition. MEME operates on frozen text, visual, and audio features, compresses variable-length modality sequences with learned-query attention pooling, and processes them with a shared hierarchical MoE backbone. Each block first applies hard modality experts for modality-specific refinement and then task-conditioned cross experts for task-aware multimodal interaction. A [TASK] token provides an explicit task anchor for routing, while modality dropping regularizes unified training. We jointly train MEME on nine affective benchmarks and evaluate all datasets using a single composite-best checkpoint without dataset-wise adaptation. Extensive experiments show that MEME outperforms strong task-specific baselines and recent large-model-based unified methods on most benchmarks, while maintaining favorable efficiency, robustness and generalization.


图片


张一楠

2024级 博士生

数据科学专业


Evaluations & Datasets Track


题目:GeoSym127K: Scalable Symbolically-verifiable Synthesis for Multimodal Geometric Reasoning                                                 Oral 

作者:Jinhao Jing, Zheng Ma, Jinwei Liang, Qiannian Zhao, Shuang Chen, Jing Yangum, Prayag Tiwari, Jingjing Bai, Benyou Wang, Por Lip Yee, Zhan Su, Lewei Lu

摘要:Large Multimodal Models (LMMs) often struggle with geometric reasoning due to visual hallucinations and a lack of mathematically precise Chain-of-Thought (CoT) data. To address this, we propose the GeoSym Engine, an automated and scalable neuro-symbolic framework. By leveraging a type-conditional grammar and an analytic SymGT Solver, it derives exact symbolic ground truths and seamlessly integrates with a robust rendering pipeline to produce high-precision geometric diagrams. Using this engine, we construct GeoSym127K, a difficulty-stratified dataset featuring 51K high-resolution images, 127K questions with symbolic ground truths, and 55K answer-verified CoT QA pairs. We also introduce GeoSym-Bench, an expert-curated suite of 511 complex samples for rigorous evaluation. Through extensive supervised fine-tuning (SFT), we demonstrate that GeoSym drives concentrated improvements specifically on diagram-dependent and multi-step geometry tasks. Our Qwen3-VL-8B model gains an absolute +22.21% on the MathVerse Vision-Only subset and reaches 61.52% (+6.19% improvement) on WeMath, mitigating long-horizon logic fragmentation and outperforming advanced closed-source models like Doubao-1.8. Furthermore, applying Reinforcement Learning with Verifiable Rewards (RLVR) via GRPO reveals that initializing from structural SFT checkpoints substantially elevates the performance ceiling over zero-shot RL. Driven by deterministic exact-match signals, this showcases the robust scaling potential of our verifiable reasoning synthesis.

图片

敬谨皓

2024级 硕士生

人工智能与机器人专业


题目:TRL-Bench: Standardizing Cross-Paradigm Representation-Level   Evaluation of Tabular Encoders    

作者:Wei Pang, Xiangru Jian, Hehan Li, Zhixuan Yu, Alex Xue, Jinyang Li, Zhengyuan Dong, Xinjian Zhao, Hao Xu, Chao Zhang, Reynold Cheng, M. Tamer Özsu, Tianshu Yu 

摘要:Tabular encoders are usually evaluated inside task-specific end-to-end pipelines, so models from different training paradigms are difficult to compare directly even when they operate on similar tabular signals. We introduce TRL-bench, a multi-granular tabular representation learning (TRL) benchmark that standardizes cross-paradigm representation-level evaluation: each encoder exports row-, column-, or table embeddings through its supported wrapper, and shared lightweight heads probe them across three suites: TRL-CTbench (column/table), TRL-Rbench (row), and TRL-DLTE (compositional Data-Lake Table Enrichment spanning all three granularities). To support this standardized setting, we release curated benchmark assets and task reformulations, including 50 OpenML tables with 123 verified targets, 16 row-pair linkage rewrites, and a 47,772-table DLTE lake derived from 1,379 parent tables. Across 20 models and 16 tasks, TRL-bench shows that once downstream conditions are standardized, encoder quality is capability-specific rather than captured by a single leaderboard. In TRL-CTbench, generic text encoders often lead on tasks with strong surface-text signal, while tabular specialists win where their pretraining objective aligns with the task. In TRL-Rbench, within-table prediction and cross-table linkage favor different training regimes, with atomic linkage performance correlating strongly with the row-matching stage of DLTE pipelines. In TRL-DLTE, the strongest pipelines combine capability-matched specialists rather than reuse a single encoder, and top end-to-end quality depends on non-additive compositional fit rather than per-stage marginal rank alone. TRL-bench provides a common protocol for measuring reusable signal in exported tabular representations under shared downstream conditions.


图片


庞威

2025级 博士生 

计算机科学专业



上述同学本科或硕士阶段就读于国内外多所知名高校,包括新加坡国立大学、复旦大学、中国科学技术大学、华南理工大学等,其中也有本院培养的学生。




指导教授简介




图片

樊继聪

助理教授

香港城市大学博士



研究领域:

人工智能、机器学习


个人简介:

樊继聪教授2018年在香港城市大学电子工程系获得博士学位,并分别于2013年和2010年在北京化工大学获得控制科学与工程硕士学位和自动化学士学位。在加入香港中文大学(深圳)之前,他是康奈尔大学的博士后。他还曾在美国威斯康星大学麦迪逊分校和香港大学担任研究职位。

樊教授的研究方向是人工智能和机器学习,他在矩阵/张量方法、聚类算法、异常检测、图学习、自动机器学习等方面做了大量研究工作,在知名学术期刊与著名国际会议上,如IEEE TSP/TNNLS/TII、NeurIPS、ICLR、ICML、CVPR、AAAI等,发表论文50余篇。 他是IEEE高级会员,目前担任期刊《Pattern Recognition》(中科院一区,CCF-B类)和《Neural Processing Letters》的副编辑,担任NeurIPS和ICML领域主席。他主持国家自然科学基金青年项目、面上项目、广东省面上项目等,获得2023年中国自动化学会自然科学奖一等奖,入选斯坦福大学/爱思唯尔2023、2024年“全球Top 2% 科学家”榜单。





图片

蒋理

助理教授

香港中文大学博士



研究领域:

计算机视觉、人工智能、机器学习


个人简介:

蒋理教授于2021年获得香港中文大学博士学位,其后在马克斯·普朗克研究所担任博士后研究员。其研究聚焦于计算机视觉与人工智能,具体方向包括三维感知与空间智能、世界模型、自动驾驶、具身智能、表征学习以及多模态学习等。其研究成果发表在CVPR、ICCV、NeurIPS、和TPAMI等顶级会议和期刊中,并多次获选口头报告与亮点论文,谷歌学术引用量超过15,000次。其关于自动驾驶运动预测的研究在2022-2024年的CVPR Waymo Open Dataset Motion Prediction Challenge中连续三届获得第一名。多次入选由斯坦福大学与爱思唯尔联合发布的全球前2%顶尖科学家年度影响力榜单,并获得国家级青年人才项目资助。


图片

刘桂良

助理教授

西蒙菲莎大学博士



研究领域:

具身智能、强化学习


个人简介:

刘桂良教授在加拿大西蒙弗雷泽大学(Simon Fraser University)计算机科学学院获得博士学位,并在加拿大滑铁卢大学(University of Waterloo)及向量研究院(Vector Institute)完成博士后研究。他的研究主要聚焦于具身智能决策与强化学习,研发了EmbodiedChain具身智能数据引擎,实现虚实迁移下的机器人操作技能生成和部署;并提出了逆约束推断强化学习模型,致力于提升强化学习控制模型的安全性和稳定性。他同时兼任跨维智能强化学习技术专家、学院具身决策实验室负责人以及河套学院研究员。刘桂良教授在包括NeurIPS、ICML、ICLR在内的国际机器学习顶级学术会议和期刊上发表论文30余篇,担任NeurIPS、ICLR领域主席,并入选深圳市“鹏城优才”及香港中文大学(深圳)“校长青年学者”计划。


图片

刘圳

助理教授

蒙特利尔大学博士



研究领域:

机器学习、生成模型、大模型后训练、三维视觉、具身智能


个人简介:

刘圳教授于2024年获得蒙特利尔大学及Mila研究所的计算机博士学位,并分别于2019年和2017从佐治亚理工学院获得计算机硕士和学士学位。2022至2024年间,他作为访问学生在德国马克思普朗克智能系统研究所从事研究工作。他的研究重点是针对1)3D建模与生成和2)视觉/语言基础模型开发有理论依据的新型机器学习方法。他在NeurIPS、ICML、ICLR、CVPR和ECCV等顶级会议上发表了多篇论文,其中数篇被选为口头报告(Oral / Spotlight)。


图片

马晨昊

助理教授

香港大学博士



研究领域:

数据管理、数据挖掘与人工智能


个人简介:

马晨昊教授主要研究方向为大规模数据管理、数据挖掘与人工智能。已在国际数据库和数据挖掘领域的重要期刊/会议(如SIGMOD、VLDB、TODS、VLDBJ、ICDE、WWW、KDD、NeurIPS、AAAI等)发表论文60余篇,其中包括以第一/通信作者身份发表的40余篇 CCF-A 长文,其中一篇代表性研究成果的论文被评为SIGMOD 2020会议的四篇最佳论文之一(~4/458),并荣获2021 ACM SIGMOD Research Highlight Award;主持国家级项目1项,省部级项目2项,CCF-蚂蚁科研基金1项(图计算专项);担任顶级会议的程序委员会领域主席/高级委员/委员(如VLDB、ICDE、KDD、WWW 、NeurIPS等)、顶级期刊(如VLDBJ、TKDE等)的审稿人。



图片

施家晟

助理教授

香港中文大学博士



研究领域:

因果分析、分布式统计理论与迁移学习、高维统计推断、自正则极限理论


个人简介:

施家晟教授于香港中文大学获得统计学博士学位,并在宾夕法尼亚大学生物统计系及费城儿童医院Policy La做博士后研究。他曾在一些统计学顶级期刊,如Annals of Statistics上发表文章。


图片

王趵翔

副教授

香港中文大学博士




研究领域:

强化学习、在线学习、学习理论


个人简介:

王趵翔教授于2014年在上海交通大学获信息安全专业工程学士学位;其后于2020年在香港中文大学计算机科学与工程系获博士学位。就读博士期间,他曾在阿尔伯塔大学和加拿大皇家银行长期访问。他的研究方向包括强化学习,在线学习,和学习理论等。他的研究成果发表在ITCS, NeurIPS, ICML, ICLR等会议。他关于The Gambler's problem的研究解决了强化学习教科书中的开放问题,并证明了强化学习中的混沌现象。



图片

王本友

助理教授

意大利帕多瓦大学博士



研究领域:

自然语言处理、信息检索、应用机器学习、量子机器学习


个人简介:

王本友教授于2022年在意大利帕多瓦大学取得博士学位。他曾是欧盟玛丽居里研究员,曾在丹麦哥本哈根大学、加拿大蒙特利尔大学、荷兰阿姆斯特丹大学、华为诺亚方舟实验室、中科院理论物理所、社科院语言所交流访问。王教授的主要研究方向为自然语言处理方面、应用机器学习、信息检索。他曾获得国际信息检索顶级会议SIGIR 2017(CCF A类会议)最佳论文提名奖,获奖论文(IRGAN)是SIGIR历史上引用最多的论文之一,是最早也是最成功的GAN在信息检索领域的应用;曾获得国际自然语言处理顶级会议NAACL 2019最佳可解释论文奖,与自然语言处理里程碑工作BERT一起同台领奖。王教授在国际顶级会议ICLR/NeurIPS/ACL/EMNLP/NAACL/SIGIR/WWW/CIKM/AAAI/IJCAI和国际顶级期刊TOIS/TOC/TCS等发表了20余篇论文。他的专著《推荐系统与深度学习》由清华大学出版社出版。他长期担任ICLR/NeurIPS/ICML审稿人。


图片

武执政

副教授

南洋理工大学博士



研究领域:

语音交互、语音生成、音频鉴伪


个人简介:

武执政教授是国家级青年人才,深圳市跨模态认知计算重点实验室副主任,华为火花奖获得者,并连续多次入选斯坦福大学“全球前2%顶尖科学家”榜单及获得最佳论文奖。武教授于南洋理工大学获得博士学位,曾任职于Meta(原Facebook)、苹果、爱丁堡大学及微软亚洲研究院等国际知名机构,从事学术研究和技术领导工作。


武教授发起了多个开源项目,包括Merlin、Amphion和Emilia,已被全球超过1000家单位采用(含OpenAI)。其中,Amphion多次登顶GitHub趋势榜,Emilia成为HuggingFace音频类最受欢迎数据集(Most Liked)。此外,他发起了语音鉴伪和语音转换国际评测,并担任2019年语音合成国际评测(Blizzard Challenge 2019)的组织者。武教授现为人工智能语音领域权威期刊IEEE/ACM TASLP、SPL等编委,并出任IEEE Spoken Language Technology Workshop 2024大会主席。


图片

严明

副教授

助理院长(本科生事务)

加州大学洛杉矶分校博士



研究领域:

分布式优化、稀疏优化、机器学习、大规模优化算法


个人简介:

严明教授于2005和2008年在中国科学技术大学获得数学本科和硕士学位,2012年在加州大学洛杉矶分校获得数学博士学位。2012-2015年,他先后在莱斯大学的计算与应用数学系和加州大学洛杉矶分校的数学系担任博士后研究员。2015-2022年,他在密西根州立大学的计算数学、科学与工程系和数学系担任助理教授(2015-2021)和副教授(2021-2022)。


他的研究方向为大规模优化算法及其在图像处理、机器学习和其他数据科学问题中的应用。他在人工智能与机器学习会议(包括ICML、ICLR、NeurIPS 和 AISTATS)以及 IEEE杂志(TPAMI、TSP) 和 SIAM 杂志(包括SIIMS、SISC、MMS)等发表文章超过50篇。他于2020年获得脸书教授奖(Facebook Faculty Award)。



图片

姚建峰

校长讲座教授

统计学学科部主任

巴黎萨克雷大学博士



研究领域:

随机矩阵理论和高维统计、高维计量经济学模型、马尔可夫链和马尔可夫过程、时间序列分析、网络数据分析、数字图像分析


个人简介:

姚建峰教授于1990年至2000年在巴黎第一大学索邦大学担任助理教授和副教授。2000年起担任雷恩第一大学数学系应用数学专业正教授。他还于1989-1994年在SUDIMAGE R&D、于2003-2004年在法国INRIA担任访问职位。2011年起在香港大学担任副教授和正教授。姚建峰教授同时也是山东大学数学学院特聘教授。


姚教授是随机矩阵理论和高维统计领域的国际领先学者。他的著作《大样本协方差矩阵和高维数据分析》(剑桥大学出版社,2015,与白志东、郑术蓉合著)是该领域的权威参考书。姚教授因其研究贡献曾多次获奖。姚教授为国际数理统计学会会士、国际统计学会会员、伯努利数理统计与概率学会科学书记和执行委员会委员。姚建峰教授担任多家著名期刊现任编委,如Journal of Multivariate Analysis、Random Matrices: Theory and Applications,曾任Bernoulli期刊编委。他曾获得评价:“对随机矩阵理论在高维数据分析中的推论方面做出了有影响力的贡献”。


图片

于天舒

助理教授

亚利桑那州立大学博士



研究领域:

科学智能、运筹智能


个人简介:

于天舒教授于2012年毕业于沈阳工业大学并取得学士学位。其后前往加拿大卡尔加里大学攻读地理信息工程专业并于2016年获得硕士学位。在此之前,他曾于2012至2014年间任飞利浦医疗算法工程师,并于2021年从亚利桑那州立大学计算机科学专业毕业取得博士学位。


于教授的主要研究兴趣涵盖多个机器学习和组合问题优化相关领域。他对利用机器学习解决传统组合问题,图学习和优化,以及在深度学习框架内寻求结构扩展特别感兴趣。循环神经网络,行列式点过程也是他开展的研究之一。他也是多个顶级会议(例如ICLR 2021, NIPS 2020, CVPR 2019-2021, ICCV 2019, ECCV 2020等)以及期刊(例如IEEE Transactions on Image Processing, IEEE Transactions on Neural Networks and Learning Systems, IEEE Transactions on Circuits and Systems for Video Technology, Pattern Recognition, Pattern Recognition Letters等)的审稿人。


图片

查宏远

校长学勤讲座教授

副院长(科研)

斯坦福大学博士



研究领域:

机器学习及应用


个人简介:

查宏远教授1984年毕业于复旦大学数学系, 并于1993年获得斯坦福大学科学计算专业博士学位。查教授于2006年至2020年任职于佐治亚理工学院计算机学院,1992年至2006年任职于宾州州立大学计算机科学与工程系,他也曾于1999年至2001年任职于 Inktomi 公司。他目前的研究方向是机器学习及应用。


查教授在计算机等相关领域的主流科技期刊和顶级学术会议上发表400多篇论文,据谷歌学术统计,截止2024年11月,谷歌H-index 93,总引用率超38349次。曾荣获多项重要学术奖项,如 Institute of Mathematics and Applications(IMA)授予的“莱斯利福克斯奖(Leslie Fox Prize)”二等奖(1991年),第34届ACM SIGIR国际信息检索会议(SIGIR 2011)最佳学生论文奖(指导教授)(2011年),第26届NeurIPS“最佳论文奖” (2013年)。

登录高考直通车APP
查看完整试题答案
好的
查看完整试题
特别声明:
1. 本站注明稿件来源为其他媒体的文/图等稿件均为转载稿,本站转载出于非商业性的教育和科研之目的,并不意味着赞同其观点或证实其内容的真实性。如转载稿涉及版权等问题,请作者在两周内速来电或来函联系,本站根据实际情况会进行下架处理。
2. 任何媒体、网站或个人未经本网协议授权不得转载、链接、转贴或以其他方式复制发表,违者本站将依法追究责任。
举报理由选择
确定
高考直通车官网

哇哦!快加入高考直通车APP,发现更多精彩内容,与学霸一起交流学习吧!