Yang Han

Liangtai Sun, Yang Han, Zihan Zhao et al.

Recently, there has been growing interest in using Large Language Models (LLMs) for scientific research. Numerous benchmarks have been proposed to evaluate the ability of LLMs for scientific research. However, current benchmarks are mostly based on pre-collected objective questions. This design suffers from data leakage problem and lacks the evaluation of subjective Q/A ability. In this paper, we propose SciEval, a comprehensive and multi-disciplinary evaluation benchmark to address these issues. Based on Bloom's taxonomy, SciEval covers four dimensions to systematically evaluate scientific research ability. In particular, we design a "dynamic" subset based on scientific principles to prevent evaluation from potential data leakage. Both objective and subjective questions are included in SciEval. These characteristics make SciEval a more effective benchmark for scientific research ability evaluation of LLMs. Comprehensive experiments on most advanced LLMs show that, although GPT-4 achieves SOTA performance compared to other LLMs, there is still substantial room for improvement, especially for dynamic questions. The codes and data are publicly available on https://github.com/OpenDFM/SciEval.

Zhida Zhang, Jie Ma, Zhan Peng et al.

The narrative quality of a video fundamentally determines its perceptual value. Although existing video generation methods can produce visually appealing content, they predominantly rely on sparse conditioning signals such as text prompts or first/last frames, which limits precise control over narrative structure and temporal pacing. In this paper, we propose SmartDirector, a framework that enhances the narrative capacity of video generation models through multiple keyframes. SmartDirector supports flexible generation scenarios including single-shot generation, multi-shot narrative synthesis, and video extension. The framework operates in two stages: Director-Gen generates a low-resolution video conditioned on the provided keyframes, and Director-SR refines the output by exploiting high-resolution keyframes as semantic anchors to recover fine-grained details. To enable robust multi-keyframe training, we construct a data pipeline that curates single-shot and multi-shot sequences from movies. Extensive experiments demonstrate that SmartDirector substantially outperforms existing state-of-the-art approaches. We will release the code to facilitate further research.

Jiaxu Wan, Xu Wang, Mengwei Xie et al.

Recent spatial intelligence approaches typically attach 3D cues to 2D reasoning pipelines or couple MLLMs with black-box reconstruction modules, leading to weak spatial consistency, limited viewpoint diversity, and evidence chains that cannot be traced back to supporting views. Frameworks for "thinking with images" (e.g., ChatGPT-o3 and DeepEyes) show that stepwise multimodal reasoning can emerge by interleaving hypothesis formation with active acquisition of visual evidence, but they do not address three key challenges in spatial Chain-of-Thought (CoT): building global space perception under strict token budgets, explicitly associating 3D hypotheses with video frames for verification, and designing spatially grounded rewards for reinforcement learning. To address these issues, we present EagleVision, a dual-stage framework for progressive spatial cognition through macro perception and micro verification. In the macro perception stage, EagleVision employs a semantics-perspective-fusion determinantal point process (SPF-DPP) to select a compact set of geometry- and semantics-aware keyframes from long videos under a fixed token budget. In the micro verification stage, we formalize spatial CoT as BEV-grounded pose querying: the agent iteratively predicts poses on a BEV plane, retrieves the nearest real frames, and is trained purely by reinforcement learning with a spatial grounding reward that scores the consistency between predicted poses and observed views. On VSI-Bench, EagleVision achieves state-of-the-art performance among open-source vision-language models, demonstrating strong and generalizable spatial understanding.

Yang Han, Zhiqiang Zou, Nan Li et al.

Astronomical outliers, such as unusual, rare or unknown types of astronomical objects or phenomena, constantly lead to the discovery of genuinely unforeseen knowledge in astronomy. More unpredictable outliers will be uncovered in principle with the increment of the coverage and quality of upcoming survey data. However, it is a severe challenge to mine rare and unexpected targets from enormous data with human inspection due to a significant workload. Supervised learning is also unsuitable for this purpose since designing proper training sets for unanticipated signals is unworkable. Motivated by these challenges, we adopt unsupervised machine learning approaches to identify outliers in the data of galaxy images to explore the paths for detecting astronomical outliers. For comparison, we construct three methods, which are built upon the k-nearest neighbors (KNN), Convolutional Auto-Encoder (CAE)+ KNN, and CAE + KNN + Attention Mechanism (attCAE KNN) separately. Testing sets are created based on the Galaxy Zoo image data published online to evaluate the performance of the above methods. Results show that attCAE KNN achieves the best recall (78%), which is 53% higher than the classical KNN method and 22% higher than CAE+KNN. The efficiency of attCAE KNN (10 minutes) is also superior to KNN (4 hours) and equal to CAE+KNN(10 minutes) for accomplishing the same task. Thus, we believe it is feasible to detect astronomical outliers in the data of galaxy images in an unsupervised manner. Next, we will apply attCAE KNN to available survey datasets to assess its applicability and reliability.

Bajian Xiang, Tingwei Guo, Xuan Chen et al.

Large Speech Language Models (LSLMs) typically operate at high token rates (tokens/s) to ensure acoustic fidelity, yet this results in sequence lengths that far exceed the underlying semantic content, incurring prohibitive inference costs. In this paper, we empirically revisit the necessity of such granular token-level processing. Through layer-wise oracle interventions, we unveil a structured redundancy hierarchy: while shallow layers encode essential acoustic details, deep layers exhibit extreme redundancy, allowing for aggressive compression. Motivated by these findings, we introduce Affinity Pooling, a training-free, similarity-based token merging mechanism. By strategically applying this method at both input and deep layers, we effectively compress speech representations without compromising semantic information. Extensive evaluations across three tasks demonstrate that our approach reduces prefilling FLOPs by 27.48\% while maintaining competitive accuracy. Practical deployment further confirms significant efficiency gains, yielding up to $\sim$1.7$\times$ memory savings and $\sim$1.1$\times$ faster time-to-first-token on long utterances. Our results challenge the necessity of fully distinct token representations, providing new perspectives on LSLM efficiency.

Lemeng Qi, Yang Han, Zhuotong Xie

This paper explores the challenges posed by nominal adjectives (NAs) in natural language processing (NLP) tasks, particularly in part-of-speech (POS) tagging. We propose treating NAs as a distinct POS tag, "JN," and investigate its impact on POS tagging, BIO chunking, and coreference resolution. Our study shows that reclassifying NAs can improve the accuracy of syntactic analysis and structural understanding in NLP. We present experimental results using Hidden Markov Models (HMMs), Maximum Entropy (MaxEnt) models, and Spacy, demonstrating the feasibility and potential benefits of this approach. Additionally we finetuned a bert model to identify the NA in untagged text.

Mingzhi Wang, Chengdong Ma, Qizhi Chen et al.

Self-play methods have demonstrated remarkable success in enhancing model capabilities across various domains. In the context of Reinforcement Learning from Human Feedback (RLHF), self-play not only boosts Large Language Model (LLM) performance but also overcomes the limitations of traditional Bradley-Terry (BT) model assumptions by finding the Nash equilibrium (NE) of a preference-based, two-player constant-sum game. However, existing methods either guarantee only average-iterate convergence, incurring high storage and inference costs, or converge to the NE of a regularized game, failing to accurately reflect true human preferences. In this paper, we introduce Magnetic Preference Optimization (MPO), a novel approach capable of achieving last-iterate convergence to the NE of the original game, effectively overcoming the limitations of existing methods. Building upon Magnetic Mirror Descent (MMD), MPO attains a linear convergence rate, making it particularly suitable for fine-tuning LLMs. To ensure our algorithm is both theoretically sound and practically viable, we present a simple yet effective implementation that adapts the theoretical insights to the RLHF setting. Empirical results demonstrate that MPO can significantly enhance the performance of LLMs, highlighting the potential of self-play methods in alignment.

Tianyi Qiu, Fanzhi Zeng, Jiaming Ji et al.

Existing alignment methods share a common topology of information flow, where reward information is collected from humans, modeled with preference learning, and used to tune language models. However, this shared topology has not been systematically characterized, nor have its alternatives been thoroughly explored, leaving the problems of low data efficiency and unreliable generalization unaddressed. As a solution, we introduce a theory of reward generalization in reinforcement learning from human feedback (RLHF), focusing on the topology of information flow at both macro and micro levels. At the macro level, we portray the RLHF information flow as an autoencoding process over behavior distributions, formalizing the RLHF objective of distributional consistency between human preference and model behavior. At the micro level, we present induced Bayesian networks to model the impact of dataset topologies on reward generalization. Combining analysis on both levels, we propose reward modeling from tree-structured preference information. It is shown to reduce reward uncertainty by up to $Θ(\log n/\log\log n)$ times compared to baselines, where $n$ is the dataset size. Validation on three NLP tasks shows that it achieves an average win rate of 65% against baselines, thus improving reward generalization for free via topology design, while reducing the amount of data requiring annotation.

Yuxin Liu, Xiong Jin, Yang Han

Label noise - incorrect labels assigned to observations - can substantially degrade the performance of supervised classifiers. This paper proposes a label noise cleaning method based on Bernoulli random sampling. We show that the mean label noise levels of subsets generated by Bernoulli random sampling containing a given observation are identically distributed for all clean observations, and identically distributed, with a different distribution, for all noisy observations. Although the mean label noise levels are not independent across observations, by introducing an independent coupling we further prove that they converge to a mixture of two well-separated distributions corresponding to clean and noisy observations. By establishing a linear model between cross-validated classification errors and label noise levels, we are able to approximate this mixture distribution and thereby separate clean and noisy observations without any prior label information. The proposed method is classifier-agnostic, theoretically justified, and demonstrates strong performance on both simulated and real datasets.

Yang Han, Ziping Wan, Lu Chen et al.

Large Language Models (LLMs) have significantly transformed our daily life and established a new paradigm in natural language processing (NLP). However, the predominant pretraining of LLMs on extensive web-based texts remains insufficient for advanced scientific discovery, particularly in chemistry. The scarcity of specialized chemistry data, coupled with the complexity of multi-modal data such as 2D graph, 3D structure and spectrum, present distinct challenges. Although several studies have reviewed Pretrained Language Models (PLMs) in chemistry, there is a conspicuous absence of a systematic survey specifically focused on chemistry-oriented LLMs. In this paper, we outline methodologies for incorporating domain-specific chemistry knowledge and multi-modal information into LLMs, we also conceptualize chemistry LLMs as agents using chemistry tools and investigate their potential to accelerate scientific research. Additionally, we conclude the existing benchmarks to evaluate chemistry ability of LLMs. Finally, we critically examine the current challenges and identify promising directions for future research. Through this comprehensive survey, we aim to assist researchers in staying at the forefront of developments in chemistry LLMs and to inspire innovative applications in the field.

Victor OK Li, Yang Han, Jacqueline CK Lam

Alzheimer's Disease (AD) affects over 55 million people globally, yet the key genetic contributors remain poorly understood. Leveraging recent advancements in genomic foundation models, we present the innovative Reverse-Gene-Finder technology, a ground-breaking neuron-to-gene-token backtracking approach in a neural network architecture to elucidate the novel causal genetic biomarkers driving AD onset. Reverse-Gene-Finder comprises three key innovations. Firstly, we exploit the observation that genes with the highest probability of causing AD, defined as the most causal genes (MCGs), must have the highest probability of activating those neurons with the highest probability of causing AD, defined as the most causal neurons (MCNs). Secondly, we utilize a gene token representation at the input layer to allow each gene (known or novel to AD) to be represented as a discrete and unique entity in the input space. Lastly, in contrast to the existing neural network architectures, which track neuron activations from the input layer to the output layer in a feed-forward manner, we develop an innovative backtracking method to track backwards from the MCNs to the input layer, identifying the Most Causal Tokens (MCTs) and the corresponding MCGs. Reverse-Gene-Finder is highly interpretable, generalizable, and adaptable, providing a promising avenue for application in other disease scenarios.

Kexin Zhu, Yang Han

The development of large language models (LLMs) has greatly promoted the progress of chatbot in multiple fields. There is an urgent need to evaluate whether LLMs can play the role of agent in housing transactions and services as well as humans. We present Real Estate Agent Large Language Model Evaluation (REAL), the first evaluation suite designed to assess the abilities of LLMs in the field of housing transactions and services. REAL comprises 5,316 high-quality evaluation entries across 4 topics: memory, comprehension, reasoning and hallucination. All these entries are organized as 14 categories to assess whether LLMs have the knowledge and ability in housing transactions and services scenario. Additionally, the REAL is used to evaluate the performance of most advanced LLMs. The experiment results indicate that LLMs still have significant room for improvement to be applied in the real estate field.

Xingya Wang, Yang Han, Yushi Liu et al.

This article quantifies the uncertainty reduction achievable for \textit{counterfactual} estimand, and cautions against potential bias when the estimand uses Digital Twins. Posed by Neyman (1923a) who showed unbiased \textit{point estimation} from designed \textit{factual} experiments is possible, \textit{counterfactual} uncertainty quantification (CUQ) remained an open challenge for about one hundred years. The $Rx: C$ \textit{counterfactual} efficacy we focus on is the ideal estimand for comparing treatment $Rx$ with control $C$, the expected outcome differential if each patient received \textit{both} $Rx$ and $C$. Enabled by our new statistical modeling principle called ETZ, we show CUQ is achievable in Randomized Controlled Trials (RCTs) with \textit{Before-and-After} Repeated Measures, common in many therapeutic areas. The CUQ we are able to achieve typically has lower variability than factual UQ. We caution against using predictors with measurement error, which violates regression assumptions and can cause \textit{attenuation} bias in estimating treatment effects. For traditional medicine and population-averaged targeted therapy, counterfactual point estimation remains unbiased. However, in both Real Human and Digital Twin approaches, estimating effects in \emph{subgroups} may suffer attenuation bias.

Yang Han, Pengyu Wang, Kai Yu et al.

Mass spectrometry (MS) plays a critical role in molecular identification, significantly advancing scientific discovery. However, structure elucidation from MS data remains challenging due to the scarcity of annotated spectra. While large-scale pretraining has proven effective in addressing data scarcity in other domains, applying this paradigm to mass spectrometry is hindered by the complexity and heterogeneity of raw spectral signals. To address this, we propose MS-BART, a unified modeling framework that maps mass spectra and molecular structures into a shared token vocabulary, enabling cross-modal learning through large-scale pretraining on reliably computed fingerprint-molecule datasets. Multi-task pretraining objectives further enhance MS-BART's generalization by jointly optimizing denoising and translation task. The pretrained model is subsequently transferred to experimental spectra through finetuning on fingerprint predictions generated with MIST, a pre-trained spectral inference model, thereby enhancing robustness to real-world spectral variability. While finetuning alleviates the distributional difference, MS-BART still suffers molecular hallucination and requires further alignment. We therefore introduce a chemical feedback mechanism that guides the model toward generating molecules closer to the reference structure. Extensive evaluations demonstrate that MS-BART achieves SOTA performance across 5/12 key metrics on MassSpecGym and NPLIB1 and is faster by one order of magnitude than competing diffusion-based methods, while comprehensive ablation studies systematically validate the model's effectiveness and robustness.

Yang Han

We introduce PM25Vision (PM25V), the largest and most comprehensive dataset to date for estimating air quality - specifically PM2.5 concentrations - from street-level images. The dataset contains over 11,114 images matched with timestamped and geolocated PM2.5 readings across 3,261 AQI monitoring stations and 11 years, significantly exceeding the scale of previous benchmarks. The spatial accuracy of this dataset has reached 5 kilometers, far exceeding the city-level accuracy of many datasets. We describe the data collection, synchronization, and cleaning pipelines, and provide baseline model performances using CNN and transformer architectures. Our dataset is publicly available.

Victor OK Li, Yang Han, Jacqueline CK Lam et al.

This study introduces Reverse-Speech-Finder (RSF), a groundbreaking neural network backtracking architecture designed to enhance Alzheimer's Disease (AD) diagnosis through speech analysis. Leveraging the power of pre-trained large language models, RSF identifies and utilizes the most probable AD-specific speech markers, addressing both the scarcity of real AD speech samples and the challenge of limited interpretability in existing models. RSF's unique approach consists of three core innovations: Firstly, it exploits the observation that speech markers most probable of predicting AD, defined as the most probable speech-markers (MPMs), must have the highest probability of activating those neurons (in the neural network) with the highest probability of predicting AD, defined as the most probable neurons (MPNs). Secondly, it utilizes a speech token representation at the input layer, allowing backtracking from MPNs to identify the most probable speech-tokens (MPTs) of AD. Lastly, it develops an innovative backtracking method to track backwards from the MPNs to the input layer, identifying the MPTs and the corresponding MPMs, and ingeniously uncovering novel speech markers for AD detection. Experimental results demonstrate RSF's superiority over traditional methods such as SHAP and Integrated Gradients, achieving a 3.5% improvement in accuracy and a 3.2% boost in F1-score. By generating speech data that encapsulates novel markers, RSF not only mitigates the limitations of real data scarcity but also significantly enhances the robustness and accuracy of AD diagnostic models. These findings underscore RSF's potential as a transformative tool in speech-based AD detection, offering new insights into AD-related linguistic deficits and paving the way for more effective non-invasive early intervention strategies.

Yang Han, Qin Guangjun, Liu Ziyuan et al.

Convolutional Neural Network (CNN) struggle to capture the multi-dimensional structural information of complex high-dimensional data, which limits their feature learning capability. This paper proposes a feature fusion method based on Topological Data Analysis (TDA) and CNN, named TDA-CNN. This method combines numerical distribution features captured by CNN with topological structure features captured by TDA to improve the feature learning and representation ability of CNN. TDA-CNN divides feature extraction into a CNN channel and a TDA channel. CNN channel extracts numerical distribution features, and the TDA channel extracts topological structure features. The two types of features are fused to form a combined feature representation, with the importance weights of each feature adaptively learned through an attention mechanism. Experimental validation on datasets such as Intel Image, Gender Images, and Chinese Calligraphy Styles by Calligraphers demonstrates that TDA-CNN improves the performance of VGG16, DenseNet121, and GoogleNet networks by 17.5%, 7.11%, and 4.45%, respectively. TDA-CNN demonstrates improved feature clustering and the ability to recognize important features. This effectively enhances the model's decision-making ability.

Liu Liu, Wenqiang Xu, Haoyuan Fu et al.

Human life is populated with articulated objects. A comprehensive understanding of articulated objects, namely appearance, structure, physics property, and semantics, will benefit many research communities. As current articulated object understanding solutions are usually based on synthetic object dataset with CAD models without physics properties, which prevent satisfied generalization from simulation to real-world applications in visual and robotics tasks. To bridge the gap, we present AKB-48: a large-scale Articulated object Knowledge Base which consists of 2,037 real-world 3D articulated object models of 48 categories. Each object is described by a knowledge graph ArtiKG. To build the AKB-48, we present a fast articulation knowledge modeling (FArM) pipeline, which can fulfill the ArtiKG for an articulated object within 10-15 minutes, and largely reduce the cost for object modeling in the real world. Using our dataset, we propose AKBNet, a novel integral pipeline for Category-level Visual Articulation Manipulation (C-VAM) task, in which we benchmark three sub-tasks, namely pose estimation, object reconstruction and manipulation. Dataset, codes, and models will be publicly available at https://liuliu66.github.io/articulationobjects/.

Rui Zhang, Yang Han, Zilong Huang et al.

This is a short technical report introducing the solution of the Team TCParser for Short-video Face Parsing Track of The 3rd Person in Context (PIC) Workshop and Challenge at CVPR 2021. In this paper, we introduce a strong backbone which is cross-window based Shuffle Transformer for presenting accurate face parsing representation. To further obtain the finer segmentation results, especially on the edges, we introduce a Feature Alignment Aggregation (FAA) module. It can effectively relieve the feature misalignment issue caused by multi-resolution feature aggregation. Benefiting from the stronger backbone and better feature aggregation, the proposed method achieves 86.9519% score in the Short-video Face Parsing track of the 3rd Person in Context (PIC) Workshop and Challenge, ranked the first place.

Yang Han, Qi Zhang, Victor O. K. Li et al.

Air pollution has long been a serious environmental health challenge, especially in metropolitan cities, where air pollutant concentrations are exacerbated by the street canyon effect and high building density. Whilst accurately monitoring and forecasting air pollution are highly crucial, existing data-driven models fail to fully address the complex interaction between air pollution and urban dynamics. Our Deep-AIR, a novel hybrid deep learning framework that combines a convolutional neural network with a long short-term memory network, aims to address this gap to provide fine-grained city-wide air pollution estimation and station-wide forecast. Our proposed framework creates 1x1 convolution layers to strengthen the learning of cross-feature spatial interaction between air pollution and important urban dynamic features, particularly road density, building density/height, and street canyon effect. Using Hong Kong and Beijing as case studies, Deep-AIR achieves a higher accuracy than our baseline models. Our model attains an accuracy of 67.6%, 77.2%, and 66.1% in fine-grained hourly estimation, 1-hr, and 24-hr air pollution forecast for Hong Kong, and an accuracy of 65.0%, 75.3%, and 63.5% for Beijing. Our saliency analysis has revealed that for Hong Kong, street canyon and road density are the best estimators for NO2, while meteorology is the best estimator for PM2.5.

Dongwei Jiang, Wubo Li, Ruixiong Zhang et al.

Building a good speech recognition system usually requires large amounts of transcribed data, which is expensive to collect. To tackle this problem, many unsupervised pre-training methods have been proposed. Among these methods, Masked Predictive Coding achieved significant improvements on various speech recognition datasets with BERT-like Masked Reconstruction loss and Transformer backbone. However, many aspects of MPC have not been fully investigated. In this paper, we conduct a further study on MPC and focus on three important aspects: the effect of pre-training data speaking style, its extension on streaming model, and how to better transfer learned knowledge from pre-training stage to downstream tasks. Experiments reveled that pre-training data with matching speaking style is more useful on downstream recognition tasks. A unified training objective with APC and MPC provided 8.46% relative error reduction on streaming model trained on HKUST. Also, the combination of target data adaption and layer-wise discriminative training helped the knowledge transfer of MPC, which achieved 3.99% relative error reduction on AISHELL over a strong baseline.