Bohan Zhang

Tao Huang, Shan You, Bohan Zhang et al.

Structural re-parameterization (Rep) methods achieve noticeable improvements on simple VGG-style networks. Despite the prevalence, current Rep methods simply re-parameterize all operations into an augmented network, including those that rarely contribute to the model's performance. As such, the price to pay is an expensive computational overhead to manipulate these unnecessary behaviors. To eliminate the above caveats, we aim to bootstrap the training with minimal cost by devising a dynamic re-parameterization (DyRep) method, which encodes Rep technique into the training process that dynamically evolves the network structures. Concretely, our proposal adaptively finds the operations which contribute most to the loss in the network, and applies Rep to enhance their representational capacity. Besides, to suppress the noisy and redundant operations introduced by Rep, we devise a de-parameterization technique for a more compact re-parameterization. With this regard, DyRep is more efficient than Rep since it smoothly evolves the given network instead of constructing an over-parameterized network. Experimental results demonstrate our effectiveness, e.g., DyRep improves the accuracy of ResNet-18 by $2.04\%$ on ImageNet and reduces $22\%$ runtime over the baseline. Code is available at: https://github.com/hunto/DyRep.

Bohan Zhang, Weidong Tang, Zhixiang Chi et al.

On-the-Fly Category Discovery (OCD) aims to recognize known classes while simultaneously discovering emerging novel categories during inference, using supervision only from known classes during offline training. Existing approaches rely either on fixed label supervision or on diffusion-based augmentations to enhance the backbone, yet none of them explicitly train the model to perform the discovery task required at test time. It is fundamentally unreasonable to expect a model optimized on limited labeled data to carry out a qualitatively different discovery objective during inference. This mismatch creates a clear optimization misalignment between the offline learning stage and the online discovery stage. In addition, prior methods often depend on hash-based encodings or severe feature compression, which further limits representational capacity. To address these issues, we propose Learning through Creation (LTC), a fully feature-based and hash-free framework that injects novel-category awareness directly into offline learning. At its core is a lightweight, online pseudo-unknown generator driven by kernel-energy minimization and entropy maximization (MKEE). Unlike previous methods that generate synthetic samples once before training, our generator evolves jointly with the model dynamics and synthesizes pseudo-novel instances on the fly at negligible cost. These samples are incorporated through a dual max-margin objective with adaptive thresholding, strengthening the model's ability to delineate and detect unknown regions through explicit creation. Extensive experiments across seven benchmarks show that LTC consistently outperforms prior work, achieving improvements ranging from 1.5 percent to 13.1 percent in all-class accuracy. The code is available at https://github.com/brandinzhang/LTC

GLM-4. 5 Team, Aohan Zeng, Xin Lv et al.

We present GLM-4.5, an open-source Mixture-of-Experts (MoE) large language model with 355B total parameters and 32B activated parameters, featuring a hybrid reasoning method that supports both thinking and direct response modes. Through multi-stage training on 23T tokens and comprehensive post-training with expert model iteration and reinforcement learning, GLM-4.5 achieves strong performance across agentic, reasoning, and coding (ARC) tasks, scoring 70.1% on TAU-Bench, 91.0% on AIME 24, and 64.2% on SWE-bench Verified. With much fewer parameters than several competitors, GLM-4.5 ranks 3rd overall among all evaluated models and 2nd on agentic benchmarks. We release both GLM-4.5 (355B parameters) and a compact version, GLM-4.5-Air (106B parameters), to advance research in reasoning and agentic AI systems. Code, models, and more information are available at https://github.com/zai-org/GLM-4.5.

Xiaokang Zhang, Sijia Luo, Bohan Zhang et al. · tsinghua

We introduce TableLLM, a robust large language model (LLM) with 8 billion parameters, purpose-built for proficiently handling tabular data manipulation tasks, whether they are embedded within documents or spreadsheets, catering to real-world office scenarios. We propose a distant supervision method for training, which comprises a reasoning process extension strategy, aiding in training LLMs to understand reasoning patterns more effectively as well as a cross-way validation strategy, ensuring the quality of the automatically generated data. To evaluate the performance of TableLLM, we have crafted benchmarks tailored to address both document and spreadsheet formats as well as constructed a well-organized evaluation pipeline capable of handling both scenarios. Thorough evaluations underscore the advantages of TableLLM when compared to various existing general-purpose and tabular data-focused LLMs. We have publicly released the model checkpoint, source code, benchmarks, and a web application for user interaction. Our codes and data are publicly available at https://github.com/TableLLM/TableLLM.

Han Li, Letian Zhu, Bohan Zhang et al.

LLM-based coding agents have shown strong performance on automated issue resolution benchmarks, yet existing evaluations largely focus on final task success, providing limited insight into how agents retrieve and use code context during problem solving. We introduce ContextBench, a process-oriented evaluation of context retrieval in coding agents. ContextBench consists of 1,136 issue-resolution tasks from 66 repositories across eight programming languages, each augmented with human-annotated gold contexts. We further implement an automated evaluation framework that tracks agent trajectories and measures context recall, precision, and efficiency throughout issue resolution. Using ContextBench, we evaluate four frontier LLMs and five coding agents. Our results show that sophisticated agent scaffolding yields only marginal gains in context retrieval ("The Bitter Lesson" of coding agents), LLMs consistently favor recall over precision, and substantial gaps exist between explored and utilized context. ContextBench augments existing end-to-end benchmarks with intermediate gold-context metrics that unbox the issue-resolution process. These contexts offer valuable intermediate signals for guiding LLM reasoning in software tasks.

Weidong Tang, Bohan Zhang, Zhixiang Chi et al.

On-the-Fly Category Discovery (OCD) requires a model, trained on an offline support set, to recognize known classes while discovering new ones from an online streaming sequence. Existing methods focus heavily on offline training. They aim to learn discriminative representations on the support set so that novel classes can be separated at test time. However, their discovery mechanism at inference is typically reduced to a single threshold. We argue that this paradigm is fundamentally flawed as OCD is not a static classification problem, but a dynamic process. The model must continuously decide 1) whether a sample belongs to a known class, 2) matches an existing novel category, or 3) should initiate a new one. Moreover, prior methods treat the support set as fixed knowledge. They do not update their decision boundaries as new evidence arrives during inference. This leads to unstable and inconsistent category formation. Our experiments confirm these issues. With properly calibrated and adaptive thresholds, substantial improvements can be achieved, even without changing the representation. Motivated by this, we propose PACO, a support-set-calibrated, tree-structured online decision framework. The framework models inference as a sequence of hierarchical decisions, including known-class routing, birth-aware novel assignment, and attach-versus-create operations over a dynamic prototype memory. Furthermore, we simulate the proxy discovery process to initialize the thresholds during offline training to align with inference. Thresholds are continuously updated during inference using mature novel prototypes. Importantly, PACO requires no heavy training and no dataset-specific tuning. It can be directly integrated into existing OCD pipelines as an inference-time module. Extensive experiments show significant improvements over SOTA baselines across seven benchmarks.

Ankit Agrawal, Jithin Garapati, Bohan Zhang

Software engineering practices for validating autonomous cyber-physical systems (e.g., Uncrewed Aerial Vehicles) remain fragmented across scenario design, simulation execution, and telemetry analysis, limiting traceability between requirements, tests, and evidence. This fragmentation reduces reproducibility, slows debugging and iteration, and hinders systematic assurance under complex and evolving environmental conditions. We present AutonomyLens, an LLM-driven framework that integrates scenario specification, simulation execution, and telemetry analysis into a unified validation workflow. AutonomyLens enables developers to translate high-level validation intent into executable, temporally evolving scenarios, automatically run simulations, and perform context-aware analysis of resulting system behavior. The framework introduces (i) a structured representation for mission-level scenarios, (ii) an automated execution pipeline, (iii) analysis mechanisms that align telemetry with scenario context to produce actionable insights, and (iv) counterfactual scenario generation that closes the loop by refining and synthesizing new test cases from observed failures. We describe the early-stage design of AutonomyLens, discuss key challenges in building integrated validation workflows for autonomous systems, and outline how such an approach can improve traceability, reproducibility, and scalability in autonomy validation.

Yukang Chen, Luozhou Wang, Wei Huang et al.

We present LongLive-2.0, an NVFP4-based parallel infrastructure throughout the full training and inference workflow of long video generation, addressing speed and memory bottlenecks. For training, we introduce sequence-parallel autoregressive (AR) training, instantiated as Balanced SP, which co-designs the efficient teacher-forcing layout with SP execution by pairing clean-history and noisy-target temporal chunks on each rank, enabling a natural teacher-forcing mask with SP-aware chunked VAE encoding. Combined with NVFP4 precision, it reduces GPU memory cost and accelerates GEMM computation during training, the proportion of which increases as video length grows. Moreover, we show that a high-quality infrastructure and dataset enable a remarkably clean training pipeline. Unlike existing Self-Forcing series methods that rely on ODE initialization and subsequent distribution matching distillation (DMD), LongLive-2.0 directly tunes a diffusion model into a long, multi-shot, interactive auto-regressive (AR) diffusion model. It can be further converted to real-time generation (4 to 2 denoising steps) with standalone LoRA weights. For inference on Blackwell GPUs, we enable W4A4 NVFP4 inference, quantize KV cache into NVFP4 for memory savings, and boost end-to-end throughput with asynchronous streaming VAE decoding. On non-Blackwell GPU architectures, we deploy SP inference to match the speed on Blackwell GPUs, while the quantized KV cache can lower inter-GPU communication of SP. Experiments show up to 2.15x speedup in training, and 1.84x in inference. LongLive-2.0-5B achieves 45.7 FPS inference while attaining strong performance on benchmarks. To our knowledge, LongLive-2.0 is the first NVFP4 training and inference system for long video generation.

Bohan Zhang, Bo Wang, Huajiang Ouyang et al.

Origami inspired architectures offer a powerful route toward lightweight, reconfigurable, and programmable robotic systems. Yet, a unified mechanics framework capable of seamlessly bridging rigid folding, elastic deformation, and stability driven transitions in compliant origami remains lacking. Here, we introduce a geometry consistent modeling framework based on discrete differential geometry (DDG) that unifies panel elasticity and crease rotation within a single variational formulation. By embedding crease panel coupling directly into a mid edge geometric discretization, the framework naturally captures rigid folding limits, distributed bending, multistability, and nonlinear dynamic snap through within one mechanically consistent structure. This unified description enables programmable control of stability and deformation across rigid and compliant regimes, allowing origami structures to transition from static folding mechanisms to active robotic modules. An implicit dynamic formulation incorporating gravity, contact, friction, and magnetic actuation further supports strongly coupled multiphysics simulations. Through representative examples spanning single fold bifurcation, deployable Miura membranes, bistable Waterbomb modules, and Kresling based crawling robots, we demonstrate how geometry driven mechanics directly informs robotic functionality. This work establishes discrete differential geometry as a foundational design language for intelligent origami robotics, enabling predictive modeling, stability programming, and mechanics guided robotic actuation within a unified computational platform.

Bohan Zhang, Xiaokang Zhang, Jing Zhang et al.

Current inference scaling methods, such as Self-consistency and Best-of-N, have proven effective in improving the accuracy of LLMs on complex reasoning tasks. However, these methods rely heavily on the quality of candidate responses and are unable to produce correct answers when all candidates are incorrect. In this paper, we propose a novel inference scaling strategy, CoT-based Synthesizer, which leverages CoT reasoning to synthesize superior answers by analyzing complementary information from multiple candidate responses, even when all candidate responses are flawed. To enable a lightweight and cost-effective implementation, we introduce an automated data generation pipeline that creates diverse training data. This allows smaller LLMs trained on this data to improve the inference accuracy of larger models, including API-based LLMs. Experimental results across four benchmark datasets with seven policy models demonstrate that our method significantly enhances performance, with gains of 11.8% for Llama3-8B and 10.3% for GPT-4o on the MATH dataset. The corresponding training data and code are publicly available on https://github.com/RUCKBReasoning/CoT-based-Synthesizer.

Sijia Luo, Xiaokang Zhang, Yuxuan Hu et al.

Reinforcement Learning (RL) has become essential for eliciting complex reasoning capabilities in Large Language Models (LLMs). However, the substantial memory overhead of storing Key-Value (KV) caches during long-horizon rollouts acts as a critical bottleneck, often prohibiting efficient training on limited hardware. While existing KV compression techniques offer a remedy for inference, directly applying them to RL training induces a severe policy mismatch, leading to catastrophic performance collapse. To address this, we introduce Sparse-RL empowers stable RL training under sparse rollouts. We show that instability arises from a fundamental policy mismatch among the dense old policy, the sparse sampler policy, and the learner policy. To mitigate this issue, Sparse-RL incorporates Sparsity-Aware Rejection Sampling and Importance-based Reweighting to correct the off-policy bias introduced by compression-induced information loss. Experimental results show that Sparse-RL reduces rollout overhead compared to dense baselines while preserving the performance. Furthermore, Sparse-RL inherently implements sparsity-aware training, significantly enhancing model robustness during sparse inference deployment.

Ruiqi Liang, Shuai Wang, Yiying Dong et al.

In the rapidly evolving field of artificial intelligence, convolutional neural networks are essential for tackling complex challenges such as machine vision and medical diagnosis. Recently, to address the challenges in processing speed and power consumption of conventional digital convolution operations, many optical components have been suggested to replace the digital convolution layer in the neural network, accelerating various machine vision tasks. Nonetheless, the analog nature of the optical convolution kernel has not been fully explored. Here, we develop a spatial frequency domain training method to create arbitrarily shaped analog convolution kernels using an optical metasurface as the convolution layer, with its receptive field largely surpassing digital convolution kernels. By employing spatial multiplexing, the multiple parallel convolution kernels with both positive and negative weights are generated under the incoherent illumination condition. We experimentally demonstrate a 98.59% classification accuracy on the MNIST dataset, with simulations showing 92.63% and 68.67% accuracy on the Fashion-MNIST and CIFAR-10 datasets with additional digital layers. This work underscores the unique advantage of analog optical convolution, offering a promising avenue to accelerate machine vision tasks, especially in edge devices.

Hussain Sadiq Abuwala, Bohan Zhang, Mushi Wang

The goal of our research was to investigate the effects of collaboration between academia and industry on Natural Language Processing (NLP). To do this, we created a pipeline to extract affiliations and citations from NLP papers and divided them into three categories: academia, industry, and hybrid (collaborations between academia and industry). Our empirical analysis found that there is a trend towards an increase in industry and academia-industry collaboration publications and that these types of publications tend to have a higher impact compared to those produced solely within academia.

Ben Yao, Qiuchi Li, Yazhou Zhang et al.

This work introduces the first benchmark for nursing value alignment, consisting of five core value dimensions distilled from international nursing codes: Altruism, Human Dignity, Integrity, Justice, and Professionalism. The benchmark comprises 1,100 real-world nursing behavior instances collected through a five-month longitudinal field study across three hospitals of varying tiers. These instances are annotated by five clinical nurses and then augmented with LLM-generated counterfactuals with reversed ethic polarity. Each original case is paired with a value-aligned and a value-violating version, resulting in 2,200 labeled instances that constitute the Easy-Level dataset. To increase adversarial complexity, each instance is further transformed into a dialogue-based format that embeds contextual cues and subtle misleading signals, yielding a Hard-Level dataset. We evaluate 23 state-of-the-art (SoTA) LLMs on their alignment with nursing values. Our findings reveal three key insights: (1) DeepSeek-V3 achieves the highest performance on the Easy-Level dataset (94.55), where Claude 3.5 Sonnet outperforms other models on the Hard-Level dataset (89.43), significantly surpassing the medical LLMs; (2) Justice is consistently the most difficult nursing value dimension to evaluate; and (3) in-context learning significantly improves alignment. This work aims to provide a foundation for value-sensitive LLMs development in clinical settings. The dataset and the code are available at https://huggingface.co/datasets/Ben012345/NurValues.

Songtao Wei, Yi Li, Bohan Zhang et al.

Parameter-Efficient Fine-Tuning (PEFT) has emerged as a practical paradigm for adapting large language models (LLMs) without updating all parameters. Most existing approaches, such as LoRA and PiSSA, rely on low-rank decompositions of weight updates. However, the low-rank assumption may restrict expressivity, particularly in task-specific adaptation scenarios where singular values are distributed relatively uniformly. To address this limitation, we propose CoSA (Compressed Sensing-Based Adaptation), a new PEFT method extended from compressed sensing theory. Instead of constraining weight updates to a low-rank subspace, CoSA expresses them through fixed random projection matrices and a compact learnable core. We provide a formal theoretical analysis of CoSA as a synthesis process, proving that weight updates can be compactly encoded into a low-dimensional space and mapped back through random projections. Extensive experimental results show that CoSA provides a principled perspective for efficient and expressive multi-scale model adaptation. Specifically, we evaluate CoSA on 10 diverse tasks, including natural language understanding and generation, employing 5 models of different scales from RoBERTa, Llama, and Qwen families. Across these settings, CoSA consistently matches or outperforms state-of-the-art PEFT methods.

Shi Qiu, Shaoyang Guo, Zhuo-Yang Song et al.

Current benchmarks for evaluating the reasoning capabilities of Large Language Models (LLMs) face significant limitations: task oversimplification, data contamination, and flawed evaluation items. These deficiencies necessitate more rigorous assessment methods. To address these limitations, we introduce PHYBench, a benchmark of 500 original physics problems ranging from high school to Physics Olympiad difficulty. PHYBench addresses data contamination through original content and employs a systematic curation pipeline to eliminate flawed items. Evaluations show that PHYBench activates more tokens and provides stronger differentiation between reasoning models compared to other baselines like AIME 2024, OlympiadBench and GPQA. Even the best-performing model, Gemini 2.5 Pro, achieves only 36.9% accuracy compared to human experts' 61.9%. To further enhance evaluation precision, we introduce the Expression Edit Distance (EED) Score for mathematical expression assessment, which improves sample efficiency by 204% over binary scoring. Moreover, PHYBench effectively elicits multi-step and multi-condition reasoning, providing a platform for examining models' reasoning robustness, preferences, and deficiencies. The benchmark results and dataset are publicly available at https://www.phybench.cn/.

Bohan Zhang, Chengke Bu, Paramveer S. Dhillon

AI writing assistants can reduce effort and improve fluency, but they may also weaken writers' sense of authorship. We study this tension with an ownership-aware co-writing editor that offers on-demand, sentence-level suggestions and tests two common design choices: persona-based coaching and style personalization. In an online study (N=176), participants completed three professional writing tasks: an email without AI help, a proposal with generic AI suggestions, and a cover letter with persona-based coaching, while half received suggestions tailored to a brief sample of their prior writing. Across the two AI-assisted tasks, psychological ownership dropped relative to unassisted writing (about 0.85-1.0 points on a 7-point scale), even as cognitive load decreased (about 0.9 points) and quality ratings stayed broadly similar overall. Persona coaching did not prevent the ownership decline. Style personalization partially restored ownership (about +0.43) and increased AI incorporation in text (+5 percentage points). We distill five design patterns: on-demand initiation, micro-suggestions, voice anchoring, audience scaffolds, and point-of-decision provenance, to guide authorship-preserving writing tools.

Lyumanshan Ye, Xiaojie Cai, Xinkai Wang et al.

This paper introduces "Interaction as Intelligence" research series, presenting a reconceptualization of human-AI relationships in deep research tasks. Traditional approaches treat interaction merely as an interface for accessing AI capabilities-a conduit between human intent and machine output. We propose that interaction itself constitutes a fundamental dimension of intelligence. As AI systems engage in extended thinking processes for research tasks, meaningful interaction transitions from an optional enhancement to an essential component of effective intelligence. Current deep research systems adopt an "input-wait-output" paradigm where users initiate queries and receive results after black-box processing. This approach leads to error cascade effects, inflexible research boundaries that prevent question refinement during investigation, and missed opportunities for expertise integration. To address these limitations, we introduce Deep Cognition, a system that transforms the human role from giving instructions to cognitive oversight-a mode of engagement where humans guide AI thinking processes through strategic intervention at critical junctures. Deep cognition implements three key innovations: (1)Transparent, controllable, and interruptible interaction that reveals AI reasoning and enables intervention at any point; (2)Fine-grained bidirectional dialogue; and (3)Shared cognitive context where the system observes and adapts to user behaviors without explicit instruction. User evaluation demonstrates that this cognitive oversight paradigm outperforms the strongest baseline across six key metrics: Transparency(+20.0%), Fine-Grained Interaction(+29.2%), Real-Time Intervention(+18.5%), Ease of Collaboration(+27.7%), Results-Worth-Effort(+8.8%), and Interruptibility(+20.7%). Evaluations on challenging research problems show 31.8% to 50.0% points of improvements over deep research systems.

Siqi Zhao, Wangyang Li, Xiru Wang et al.

Machine learning (ML) and deep learning (DL) techniques have been widely applied to analyze electroencephalography (EEG) signals for disease diagnosis and brain-computer interfaces (BCI). The integration of multimodal data has been shown to enhance the accuracy of ML and DL models. Combining EEG with other modalities can improve clinical decision-making by addressing complex tasks in clinical populations. This systematic literature review explores the use of multimodal EEG data in ML and DL models for clinical applications. A comprehensive search was conducted across PubMed, Web of Science, and Google Scholar, yielding 16 relevant studies after three rounds of filtering. These studies demonstrate the application of multimodal EEG data in addressing clinical challenges, including neuropsychiatric disorders, neurological conditions (e.g., seizure detection), neurodevelopmental disorders (e.g., autism spectrum disorder), and sleep stage classification. Data fusion occurred at three levels: signal, feature, and decision levels. The most commonly used ML models were support vector machines (SVM) and decision trees. Notably, 11 out of the 16 studies reported improvements in model accuracy with multimodal EEG data. This review highlights the potential of multimodal EEG-based ML models in enhancing clinical diagnostics and problem-solving.

Bohan Zhang, Yixin Wang, Paramveer S. Dhillon

In this paper, we examine the collaborative dynamics between humans and language models (LMs), where the interactions typically involve LMs proposing text segments and humans editing or responding to these proposals. Productive engagement with LMs in such scenarios necessitates that humans discern effective text-based interaction strategies, such as editing and response styles, from historical human-LM interactions. This objective is inherently causal, driven by the counterfactual `what-if' question: how would the outcome of collaboration change if humans employed a different text editing/refinement strategy? A key challenge in answering this causal inference question is formulating an appropriate causal estimand: the conventional average treatment effect (ATE) estimand is inapplicable to text-based treatments due to their high dimensionality. To address this concern, we introduce a new causal estimand -- Incremental Stylistic Effect (ISE) -- which characterizes the average impact of infinitesimally shifting a text towards a specific style, such as increasing formality. We establish the conditions for the non-parametric identification of ISE. Building on this, we develop CausalCollab, an algorithm designed to estimate the ISE of various interaction strategies in dynamic human-LM collaborations. Our empirical investigations across three distinct human-LM collaboration scenarios reveal that CausalCollab effectively reduces confounding and significantly improves counterfactual estimation over a set of competitive baselines.

Bohan Zhang, Zihao Wang, Hengyu Fu et al.

In deep learning, a central issue is to understand how neural networks efficiently learn high-dimensional features. To this end, we explore the gradient descent learning of a general Gaussian Multi-index model $f(\boldsymbol{x})=g(\boldsymbol{U}\boldsymbol{x})$ with hidden subspace $\boldsymbol{U}\in \mathbb{R}^{r\times d}$, which is the canonical setup to study representation learning. We prove that under generic non-degenerate assumptions on the link function, a standard two-layer neural network trained via layer-wise gradient descent can agnostically learn the target with $o_d(1)$ test error using $\widetilde{\mathcal{O}}(d)$ samples and $\widetilde{\mathcal{O}}(d^2)$ time. The sample and time complexity both align with the information-theoretic limit up to leading order and are therefore optimal. During the first stage of gradient descent learning, the proof proceeds via showing that the inner weights can perform a power-iteration process. This process implicitly mimics a spectral start for the whole span of the hidden subspace and eventually eliminates finite-sample noise and recovers this span. It surprisingly indicates that optimal results can only be achieved if the first layer is trained for more than $\mathcal{O}(1)$ steps. This work demonstrates the ability of neural networks to effectively learn hierarchical functions with respect to both sample and time efficiency.

Bohan Zhang, Yiyi Miao, Taoyu Wu et al.

A structured understanding of intraoral 3D scans is essential for digital orthodontics. However, existing deep-learning approaches rely heavily on modality-specific training, large annotated datasets, and controlled scanning conditions, which limit generalization across devices and hinder deployment in real clinical workflows. Moreover, raw intraoral meshes exhibit substantial variation in arch pose, incomplete geometry caused by occlusion or tooth contact, and a lack of texture cues, making unified semantic interpretation highly challenging. To address these limitations, we propose ArchMap, a training-free and knowledge-guided framework for robust structured dental understanding. ArchMap first introduces a geometry-aware arch-flattening module that standardizes raw 3D meshes into spatially aligned, continuity-preserving multi-view projections. We then construct a Dental Knowledge Base (DKB) encoding hierarchical tooth ontology, dentition-stage policies, and clinical semantics to constrain the symbolic reasoning space. We validate ArchMap on 1060 pre-/post-orthodontic cases, demonstrating robust performance in tooth counting, anatomical partitioning, dentition-stage classification, and the identification of clinical conditions such as crowding, missing teeth, prosthetics, and caries. Compared with supervised pipelines and prompted VLM baselines, ArchMap achieves higher accuracy, reduced semantic drift, and superior stability under sparse or artifact-prone conditions. As a fully training-free system, ArchMap demonstrates that combining geometric normalization with ontology-guided multimodal reasoning offers a practical and scalable solution for the structured analysis of 3D intraoral scans in modern digital orthodontics.

Xiaoyuan Li, Xinru Xue, Bohan Zhang et al.

Brain-computer interface (BCI) technology enables direct communication between the brain and external devices through electroencephalography (EEG) signals. However, existing decoding models often mix common and personalized components, leading to interference from individual variability that limits cross-subject decoding performance. To address this issue, this paper proposes a system filter that extends the concept of signal filtering to the system level. The method expands a system into its spectral representation, selectively removes unnecessary components, and reconstructs the system from the retained target components, thereby achieving explicit system-level decomposition and filtering. We further integrate the system filter into a Cross-Subject Decoding framework based on the System Filter (CSD-SF) and evaluate it on the four-class motor imagery (MI) task of the BCIC IV 2a dataset. Personalized models are transformed into relation spectrums, and statistical testing across subjects is used to remove personalized components. The remaining stable relations, representing common components across subjects, are then used to construct a common model for cross-subject decoding. Experimental results show an average improvement of 3.28% in decoding accuracy over baseline methods, demonstrating that the proposed system filter effectively isolates stable common components and enhances model robustness and generalizability in cross-subject EEG decoding.

Yachuan Liu, Xiaochun Wei, Lin Shi et al.

Large language models (LLMs) face significant challenges in ex-ante reasoning, where analysis, inference, or predictions must be made without access to information from future events. Even with explicit prompts enforcing temporal cutoffs, LLMs often generate outputs influenced by internalized knowledge of events beyond the specified cutoff. This paper introduces a novel task and benchmark designed to evaluate the ability of LLMs to reason while adhering to such temporal constraints. The benchmark includes a variety of tasks: stock prediction, Wikipedia event prediction, scientific publication prediction, and Question Answering (QA), designed to assess factual knowledge under temporal cutoff constraints. We use leakage rate to quantify models' reliance on future information beyond cutoff timestamps. Experimental results reveal that LLMs struggle to consistently adhere to temporal cutoffs across common prompting strategies and tasks, demonstrating persistent challenges in ex-ante reasoning. This benchmark provides a potential evaluation framework to advance the development of LLMs' temporal reasoning ability for time-sensitive applications.

Jiaxin Liu, Xiaoqian Jiang, Xiang Li et al.

Fairness has been a significant challenge in graph neural networks (GNNs) since degree biases often result in un-equal prediction performance among nodes with varying degrees. Existing GNN models focus on prediction accuracy, frequently overlooking fairness across different degree groups. To addressthis issue, we propose a novel GNN framework, namely Fairness- Aware Asymmetric Contrastive Ensemble (FairACE), which inte-grates asymmetric contrastive learning with adversarial training to improve degree fairness. FairACE captures one-hop local neighborhood information and two-hop monophily similarity to create fairer node representations and employs a degree fairness regulator to balance performance between high-degree and low-degree nodes. During model training, a novel group-balanced fairness loss is proposed to minimize classification disparities across degree groups. In addition, we also propose a novel fairness metric, the Accuracy Distribution Gap (ADG), which can quantitatively assess and ensure equitable performance across different degree-based node groups. Experimental results on both synthetic and real-world datasets demonstrate that FairACE significantly improves degree fairness metrics while maintaining competitive accuracy in comparison to the state-of-the-art GNN models.

Bohan Zhang, Yixin Wang, Paramveer S. Dhillon

This paper introduces a novel causal framework for multi-stage decision-making in natural language action spaces where outcomes are only observed after a sequence of actions. While recent approaches like Proximal Policy Optimization (PPO) can handle such delayed-reward settings in high-dimensional action spaces, they typically require multiple models (policy, value, and reward) and substantial training data. Our approach employs Q-learning to estimate Dynamic Treatment Regimes (DTR) through a single model, enabling data-efficient policy learning via gradient ascent on language embeddings. A key technical contribution of our approach is a decoding strategy that translates optimized embeddings back into coherent natural language. We evaluate our approach on mental health intervention, hate speech countering, and sentiment transfer tasks, demonstrating significant improvements over competitive baselines across multiple metrics. Notably, our method achieves superior transfer strength while maintaining content preservation and fluency, as validated through human evaluation. Our work provides a practical foundation for learning optimal policies in complex language tasks where training data is limited.

Zeyao Ma, Bohan Zhang, Jing Zhang et al.

We introduce SpreadsheetBench, a challenging spreadsheet manipulation benchmark exclusively derived from real-world scenarios, designed to immerse current large language models (LLMs) in the actual workflow of spreadsheet users. Unlike existing benchmarks that rely on synthesized queries and simplified spreadsheet files, SpreadsheetBench is built from 912 real questions gathered from online Excel forums, which reflect the intricate needs of users. The associated spreadsheets from the forums contain a variety of tabular data such as multiple tables, non-standard relational tables, and abundant non-textual elements. Furthermore, we propose a more reliable evaluation metric akin to online judge platforms, where multiple spreadsheet files are created as test cases for each instruction, ensuring the evaluation of robust solutions capable of handling spreadsheets with varying values. Our comprehensive evaluation of various LLMs under both single-round and multi-round inference settings reveals a substantial gap between the state-of-the-art (SOTA) models and human performance, highlighting the benchmark's difficulty.

Yachuan Liu, Bohan Zhang, Qiaozhu Mei et al.

Recent work has shown that standard training via empirical risk minimization (ERM) can produce models that achieve high accuracy on average but low accuracy on underrepresented groups due to the prevalence of spurious features. A predominant approach to tackle this group robustness problem minimizes the worst group error (akin to a minimax strategy) on the training data, hoping it will generalize well on the testing data. However, this is often suboptimal, especially when the out-of-distribution (OOD) test data contains previously unseen groups. Inspired by ideas from the information retrieval and learning-to-rank literature, this paper first proposes to use Discounted Cumulative Gain (DCG) as a metric of model quality for facilitating better hyperparameter tuning and model selection. Being a ranking-based metric, DCG weights multiple poorly-performing groups (instead of considering just the group with the worst performance). As a natural next step, we build on our results to propose a ranking-based training method called Discounted Rank Upweighting (DRU), which differentially reweights a ranked list of poorly-performing groups in the training data to learn models that exhibit strong OOD performance on the test data. Results on several synthetic and real-world datasets highlight the superior generalization ability of our group-ranking-based (akin to soft-minimax) approach in selecting and learning models that are robust to group distributional shifts.

Nima Sadri, Bohan Zhang, Bihan Liu

Creating abstractive summaries from meeting transcripts has proven to be challenging due to the limited amount of labeled data available for training neural network models. Moreover, Transformer-based architectures have proven to beat state-of-the-art models in summarizing news data. In this paper, we utilize a Transformer-based Pointer Generator Network to generate abstract summaries for meeting transcripts. This model uses 2 LSTMs as an encoder and a decoder, a Pointer network which copies words from the inputted text, and a Generator network to produce out-of-vocabulary words (hence making the summary abstractive). Moreover, a coverage mechanism is used to avoid repetition of words in the generated summary. First, we show that training the model on a news summary dataset and using zero-shot learning to test it on the meeting dataset proves to produce better results than training it on the AMI meeting dataset. Second, we show that training this model first on out-of-domain data, such as the CNN-Dailymail dataset, followed by a fine-tuning stage on the AMI meeting dataset is able to improve the performance of the model significantly. We test our model on a testing set from the AMI dataset and report the ROUGE-2 score of the generated summary to compare with previous literature. We also report the Factual score of our summaries since it is a better benchmark for abstractive summaries since the ROUGE-2 score is limited to measuring word-overlaps. We show that our improved model is able to improve on previous models by at least 5 ROUGE-2 scores, which is a substantial improvement. Also, a qualitative analysis of the summaries generated by our model shows that these summaries and human-readable and indeed capture most of the important information from the transcripts.

Yinglin Duan, Tianyang Shi, Zhengxia Zou et al.

Motion completion is a challenging and long-discussed problem, which is of great significance in film and game applications. For different motion completion scenarios (in-betweening, in-filling, and blending), most previous methods deal with the completion problems with case-by-case designs. In this work, we propose a simple but effective method to solve multiple motion completion problems under a unified framework and achieves a new state of the art accuracy under multiple evaluation settings. Inspired by the recent great success of attention-based models, we consider the completion as a sequence to sequence prediction problem. Our method consists of two modules - a standard transformer encoder with self-attention that learns long-range dependencies of input motions, and a trainable mixture embedding module that models temporal information and discriminates key-frames. Our method can run in a non-autoregressive manner and predict multiple missing frames within a single forward propagation in real time. We finally show the effectiveness of our method in music-dance applications.