Jiawei Gu

Zhaochen Su, Linjie Li, Mingyang Song et al. · microsoft-research

While humans can flexibly leverage interactive visual cognition for complex problem-solving, enabling Large Vision-Language Models (LVLMs) to learn similarly adaptive behaviors with visual tools remains challenging. A significant hurdle is the current lack of standardized infrastructure, which hinders integrating diverse tools, generating rich interaction data, and training robust agents effectively. To address these gaps, we introduce OpenThinkIMG, the first open-source, comprehensive end-to-end framework for tool-augmented LVLMs. It features standardized vision tool interfaces, scalable trajectory generation for policy initialization, and a flexible training environment. Furthermore, considering supervised fine-tuning (SFT) on static demonstrations offers limited policy generalization for dynamic tool invocation, we propose a novel reinforcement learning (RL) framework V-ToolRL to train LVLMs to learn adaptive policies for invoking external vision tools. V-ToolRL enables LVLMs to autonomously discover optimal tool-usage strategies by directly optimizing for task success using feedback from tool interactions. We empirically validate V-ToolRL on challenging chart reasoning tasks. Our RL-trained agent, built upon a Qwen2-VL-2B, significantly outperforms its SFT-initialized counterpart (+28.83 points) and surpasses established supervised tool-learning baselines like Taco and CogCom by an average of +12.7 points. Notably, it also surpasses prominent closed-source models like GPT-4.1 by +8.68 accuracy points. We hope OpenThinkIMG can serve as a foundational framework for advancing dynamic, tool-augmented visual reasoning, helping the community develop AI agents that can genuinely "think with images".

Jiawei Gu, Zacc Yang, Chuanghao Ding et al.

Large Language Models (LLMs) excel in diverse tasks but often underperform in specialized fields due to limited domain-specific or proprietary corpus. Continual pre-training (CPT) enhances LLM capabilities by imbuing new domain-specific or proprietary knowledge while replaying general corpus to prevent catastrophic forgetting. The data mixture ratio of general corpus and domain-specific corpus, however, has been chosen heuristically, leading to sub-optimal training efficiency in practice. In this context, we attempt to re-visit the scaling behavior of LLMs under the hood of CPT, and discover a power-law relationship between loss, mixture ratio, and training tokens scale. We formalize the trade-off between general and domain-specific capabilities, leading to a well-defined Critical Mixture Ratio (CMR) of general and domain data. By striking the balance, CMR maintains the model's general ability and achieves the desired domain transfer, ensuring the highest utilization of available resources. Considering the balance between efficiency and effectiveness, CMR can be regarded as the optimal mixture ratio. Through extensive experiments, we ascertain the predictability of CMR, propose CMR scaling law and have substantiated its generalization. These findings offer practical guidelines for optimizing LLM training in specialized domains, ensuring both general and domain-specific performance while efficiently managing training resources.

Tingchen Fu, Jiawei Gu, Yafu Li et al.

Instruction-following is essential for aligning large language models (LLMs) with user intent. While recent reasoning-oriented models exhibit impressive performance on complex mathematical problems, their ability to adhere to natural language instructions remains underexplored. In this work, we introduce MathIF, a dedicated benchmark for evaluating instruction-following in mathematical reasoning tasks. Our empirical analysis reveals a consistent tension between scaling up reasoning capacity and maintaining controllability, as models that reason more effectively often struggle to comply with user directives. We find that models tuned on distilled long chains-of-thought or trained with reasoning-oriented reinforcement learning often degrade in instruction adherence, especially when generation length increases. Furthermore, we show that even simple interventions can partially recover obedience, though at the cost of reasoning performance. These findings highlight a fundamental tension in current LLM training paradigms and motivate the need for more instruction-aware reasoning models. We release the code and data at https://github.com/TingchenFu/MathIF.

Mingyang Song, Haoyu Sun, Jiawei Gu et al.

When humans face problems beyond their immediate capabilities, they rely on tools, providing a promising paradigm for improving visual reasoning in multimodal large language models (MLLMs). Effective reasoning, therefore, hinges on knowing which tools to use, when to invoke them, and how to compose them over multiple steps, even when faced with new tools or new tasks. We introduce \textbf{AdaReasoner}, a family of multimodal models that learn tool use as a general reasoning skill rather than as tool-specific or explicitly supervised behavior. AdaReasoner is enabled by (i) a scalable data curation pipeline exposing models to long-horizon, multi-step tool interactions; (ii) Tool-GRPO, a reinforcement learning algorithm that optimizes tool selection and sequencing based on end-task success; and (iii) an adaptive learning mechanism that dynamically regulates tool usage. Together, these components allow models to infer tool utility from task context and intermediate outcomes, enabling coordination of multiple tools and generalization to unseen tools. Empirically, AdaReasoner exhibits strong tool-adaptive and generalization behaviors: it autonomously adopts beneficial tools, suppresses irrelevant ones, and adjusts tool usage frequency based on task demands, despite never being explicitly trained to do so. These capabilities translate into state-of-the-art performance across challenging benchmarks, improving the 7B base model by +24.9\% on average and surpassing strong proprietary systems such as GPT-5 on multiple tasks, including VSP and Jigsaw.

Jiawei Gu, Xuhui Jiang, Zhichao Shi et al.

Accurate and consistent evaluation is crucial for decision-making across numerous fields, yet it remains a challenging task due to inherent subjectivity, variability, and scale. Large Language Models (LLMs) have achieved remarkable success across diverse domains, leading to the emergence of "LLM-as-a-Judge," where LLMs are employed as evaluators for complex tasks. With their ability to process diverse data types and provide scalable, cost-effective, and consistent assessments, LLMs present a compelling alternative to traditional expert-driven evaluations. However, ensuring the reliability of LLM-as-a-Judge systems remains a significant challenge that requires careful design and standardization. This paper provides a comprehensive survey of LLM-as-a-Judge, addressing the core question: How can reliable LLM-as-a-Judge systems be built? We explore strategies to enhance reliability, including improving consistency, mitigating biases, and adapting to diverse assessment scenarios. Additionally, we propose methodologies for evaluating the reliability of LLM-as-a-Judge systems, supported by a novel benchmark designed for this purpose. To advance the development and real-world deployment of LLM-as-a-Judge systems, we also discussed practical applications, challenges, and future directions. This survey serves as a foundational reference for researchers and practitioners in this rapidly evolving field.

Haoyu Sun, Huichen Will Wang, Jiawei Gu et al. · uw

Front-end engineering involves a complex workflow where engineers conceptualize designs, translate them into code, and iteratively refine the implementation. While recent benchmarks primarily focus on converting visual designs to code, we present FullFront, a benchmark designed to evaluate Multimodal Large Language Models (MLLMs) \textbf{across the full front-end development pipeline}. FullFront assesses three fundamental tasks that map directly to the front-end engineering pipeline: Webpage Design (conceptualization phase), Webpage Perception QA (comprehension of visual organization and elements), and Webpage Code Generation (implementation phase). Unlike existing benchmarks that use either scraped websites with bloated code or oversimplified LLM-generated HTML, FullFront employs a novel, two-stage process to transform real-world webpages into clean, standardized HTML while maintaining diverse visual designs and avoiding copyright issues. Extensive testing of state-of-the-art MLLMs reveals significant limitations in page perception, code generation (particularly for image handling and layout), and interaction implementation. Our results quantitatively demonstrate performance disparities across models and tasks, and highlight a substantial gap between current MLLM capabilities and human expert performance in front-end engineering. The FullFront benchmark and code are available in https://github.com/Mikivishy/FullFront.

Jiawei Gu, Yunzhuo Hao, Huichen Will Wang et al.

Multimodal reasoning requires iterative coordination between language and vision, yet it remains unclear what constitutes a meaningful interleaved chain of thought. We posit that text and image thoughts should function as complementary rather than isomorphic modalities that mutually advance reasoning. Guided by this principle, we build ThinkMorph, a unified model fine-tuned on approximately 24K high-quality interleaved reasoning traces spanning tasks with varying visual engagement. ThinkMorph learns to generate progressive text-image reasoning steps that concretely manipulate visual content while maintaining coherent verbal logic. It delivers large gains on vision-centric benchmarks (averaging 34.7 percent over the base model) and generalizes to out-of-domain tasks, matching or surpassing larger and proprietary VLMs. Beyond performance, ThinkMorph exhibits emergent multimodal intelligence, including unseen visual manipulation skills, adaptive switching between reasoning modes, and better test-time scaling through diversified multimodal thoughts. These findings suggest promising directions for characterizing the emergent capabilities of unified models for multimodal reasoning.

Yangyang Wang, Jiawei Gu, Li Long et al.

Accurate demand estimation is critical for the retail business in guiding the inventory and pricing policies of perishable products. However, it faces fundamental challenges from censored sales data during stockouts, where unobserved demand creates systemic policy biases. Existing datasets lack the temporal resolution and annotations needed to address this censoring effect. To fill this gap, we present FreshRetailNet-50K, the first large-scale benchmark for censored demand estimation. It comprises 50,000 store-product time series of detailed hourly sales data from 898 stores in 18 major cities, encompassing 863 perishable SKUs meticulously annotated for stockout events. The hourly stock status records unique to this dataset, combined with rich contextual covariates, including promotional discounts, precipitation, and temporal features, enable innovative research beyond existing solutions. We demonstrate one such use case of two-stage demand modeling: first, we reconstruct the latent demand during stockouts using precise hourly annotations. We then leverage the recovered demand to train robust demand forecasting models in the second stage. Experimental results show that this approach achieves a 2.73% improvement in prediction accuracy while reducing the systematic demand underestimation from 7.37% to near-zero bias. With unprecedented temporal granularity and comprehensive real-world information, FreshRetailNet-50K opens new research directions in demand imputation, perishable inventory optimization, and causal retail analytics. The unique annotation quality and scale of the dataset address long-standing limitations in retail AI, providing immediate solutions and a platform for future methodological innovation. The data (https://huggingface.co/datasets/Dingdong-Inc/FreshRetailNet-50K) and code (https://github.com/Dingdong-Inc/frn-50k-baseline}) are openly released.

Yunzhuo Hao, Jiawei Gu, Huichen Will Wang et al. · microsoft-research, uw

The ability to organically reason over and with both text and images is a pillar of human intelligence, yet the ability of Multimodal Large Language Models (MLLMs) to perform such multimodal reasoning remains under-explored. Existing benchmarks often emphasize text-dominant reasoning or rely on shallow visual cues, failing to adequately assess integrated visual and textual reasoning. We introduce EMMA (Enhanced MultiModal reAsoning), a benchmark targeting organic multimodal reasoning across mathematics, physics, chemistry, and coding. EMMA tasks demand advanced cross-modal reasoning that cannot be addressed by reasoning independently in each modality, offering an enhanced test suite for MLLMs' reasoning capabilities. Our evaluation of state-of-the-art MLLMs on EMMA reveals significant limitations in handling complex multimodal and multi-step reasoning tasks, even with advanced techniques like Chain-of-Thought prompting and test-time compute scaling underperforming. These findings underscore the need for improved multimodal architectures and training paradigms to close the gap between human and model reasoning in multimodality.

Fanqing Meng, Lingxiao Du, Zijian Wu et al.

Language-model agents are increasingly used as persistent coworkers that assist users across multiple working days. During such workflows, the surrounding environment may change independently of the agent: new emails arrive, calendar entries shift, knowledge-base records are updated, and evidence appears across images, scanned PDFs, audio, video, and spreadsheets. Existing benchmarks do not adequately evaluate this setting because they typically run within a single static episode and remain largely text-centric. We introduce \bench{}, a benchmark for coworker agents built around multi-turn multi-day tasks, a stateful sandboxed service environment whose state evolves between turns, and rule-based verification. The current release contains 100 tasks across 13 professional scenarios, executed against five stateful sandboxed services (filesystem, email, calendar, knowledge base, spreadsheet) and scored by 1537 deterministic Python checkers over post-execution service state; no LLM-as-judge is invoked during scoring. We benchmark seven frontier agent systems. The strongest model reaches 75.8 weighted score, but the best strict Task Success is only 20.0\%, indicating that partial progress is common while complete end-to-end workflow completion remains rare. Turn-level analysis shows that performance drops after the first exogenous environment update, highlighting adaptation to changing state as a key open challenge. We release the benchmark, evaluation harness, and construction pipeline to support reproducible coworker-agent evaluation.

Chenkun Zhang, Jiawei Gu, Lei-Hong Zhang

We propose NEP_MiniMax, a novel computational method for solving nonlinear eigenvalue problems (NEPs) $T(λ)\mathbf{u}= 0$ on compact continua $Ω\subset \mathbb{C}$. The method combines two key components: (1) a rational minimax approximation scheme where the {m-d-Lawson} algorithm constructs a minimax rational approximation for the vector-valued function from $T(x)$'s split form, yielding a matrix-valued rational approximation $R^*(x) = P^*(x)/q^*(x) \approx T(x)$, and (2) a structure-exploiting linearization technique. The minimax approximation guarantees uniform accuracy while generally keeping $R^*(x)$ pole-free in $Ω$. Eigenpairs are then computed by solving a polynomial eigenvalue problem $P^*(λ) \mathbf{u}= 0$ via a strong linearization that exactly preserves eigenvalue multiplicities. Numerical experiments on benchmarks from the NLEVP collection demonstrate competitiveness with state-of-the-art methods (e.g., Beyn, NLEIGS, SV-AAA) in efficiency and accuracy, with theoretical error bounds directly relating eigenpair approximations to the rational approximation quality.

Linjie Li, Mahtab Bigverdi, Jiawei Gu et al.

Spatial cognition is essential for human intelligence, enabling problem-solving through visual simulations rather than solely relying on verbal reasoning. However, existing AI benchmarks primarily assess verbal reasoning, neglecting the complexities of non-verbal, multi-step visual simulation. We introduce STARE(Spatial Transformations and Reasoning Evaluation), a benchmark designed to rigorously evaluate multimodal large language models on tasks better solved through multi-step visual simulation. STARE features 4K tasks spanning foundational geometric transformations (2D and 3D), integrated spatial reasoning (cube net folding and tangram puzzles), and real-world spatial reasoning (perspective and temporal reasoning), reflecting practical cognitive challenges like object assembly, mechanical diagram interpretation, and everyday spatial navigation. Our evaluations show that models excel at reasoning over simpler 2D transformations, but perform close to random chance on more complex tasks like 3D cube net folding and tangram puzzles that require multi-step visual simulations. Humans achieve near-perfect accuracy but take considerable time (up to 28.9s) on complex tasks, significantly speeding up (down by 7.5 seconds on average) with intermediate visual simulations. In contrast, models exhibit inconsistent performance gains from visual simulations, improving on most tasks but declining in specific cases like tangram puzzles (GPT-4o, o1) and cube net folding (Claude-3.5, Gemini-2.0 Flash), indicating that models may not know how to effectively leverage intermediate visual information.

Jiawei Gu, Ziting Xian, Yuanzhen Xie et al.

Large language models (LLMs) achieve strong performance on plain text tasks but underperform on structured data like tables and databases. Potential challenges arise from their underexposure during pre-training and rigid text-to-structure transfer mechanisms. Unlike humans who seamlessly apply learned patterns across data modalities, LLMs struggle to infer implicit relationships embedded in tabular formats, especially in the absence of explicit structural guidance. To bridge this cognitive gap, we introduce Contrastive Retrieval-Augmented Generation on Experience (CoRE), a framework that builds experience memory representations and enhances generalization through contrastive In-Context Learning (ICL) to simulate human-like knowledge transfer. Experiments on Text-to-SQL and TableQA show CoRE significantly improves performance, achieving average gains of 3.44% and 4.24%, with up to 17.2% on challenging tasks. Our Monte Carlo Tree Search (MCTS)-generated Experience Memory expands training data 8-9x, enhancing diversity and domain coverage. This training-free and continual method propels LLMs toward structured knowledge expertise.

Ziyue Qiao, Qianyi Cai, Hao Dong et al.

This paper addresses the challenge of graph domain adaptation on evolving, multiple out-of-distribution (OOD) graphs. Conventional graph domain adaptation methods are confined to single-step adaptation, making them ineffective in handling continuous domain shifts and prone to catastrophic forgetting. This paper introduces the Graph Continual Adaptive Learning (GCAL) method, designed to enhance model sustainability and adaptability across various graph domains. GCAL employs a bilevel optimization strategy. The "adapt" phase uses an information maximization approach to fine-tune the model with new graph domains while re-adapting past memories to mitigate forgetting. Concurrently, the "generate memory" phase, guided by a theoretical lower bound derived from information bottleneck theory, involves a variational memory graph generation module to condense original graphs into memories. Extensive experimental evaluations demonstrate that GCAL substantially outperforms existing methods in terms of adaptability and knowledge retention.

Yunchao Liu, Ha Dong, Xin Wang et al.

While deep learning has revolutionized computer-aided drug discovery, the AI community has predominantly focused on model innovation and placed less emphasis on establishing best benchmarking practices. We posit that without a sound model evaluation framework, the AI community's efforts cannot reach their full potential, thereby slowing the progress and transfer of innovation into real-world drug discovery. Thus, in this paper, we seek to establish a new gold standard for small molecule drug discovery benchmarking, WelQrate. Specifically, our contributions are threefold: WelQrate Dataset Collection - we introduce a meticulously curated collection of 9 datasets spanning 5 therapeutic target classes. Our hierarchical curation pipelines, designed by drug discovery experts, go beyond the primary high-throughput screen by leveraging additional confirmatory and counter screens along with rigorous domain-driven preprocessing, such as Pan-Assay Interference Compounds (PAINS) filtering, to ensure the high-quality data in the datasets; WelQrate Evaluation Framework - we propose a standardized model evaluation framework considering high-quality datasets, featurization, 3D conformation generation, evaluation metrics, and data splits, which provides a reliable benchmarking for drug discovery experts conducting real-world virtual screening; Benchmarking - we evaluate model performance through various research questions using the WelQrate dataset collection, exploring the effects of different models, dataset quality, featurization methods, and data splitting strategies on the results. In summary, we recommend adopting our proposed WelQrate as the gold standard in small molecule drug discovery benchmarking. The WelQrate dataset collection, along with the curation codes, and experimental scripts are all publicly available at WelQrate.org.

Jiawei Gu, Ziyue Qiao, Zechao Li

The task of graph-level out-of-distribution (OOD) detection is crucial for deploying graph neural networks in real-world settings. In this paper, we observe a significant difference in the relationship between the largest and second-largest eigenvalues of the Laplacian matrix for in-distribution (ID) and OOD graph samples: \textit{OOD samples often exhibit anomalous spectral gaps (the difference between the largest and second-largest eigenvalues)}. This observation motivates us to propose SpecGap, an effective post-hoc approach for OOD detection on graphs. SpecGap adjusts features by subtracting the component associated with the second-largest eigenvalue, scaled by the spectral gap, from the high-level features (i.e., $\mathbf{X}-\left(λ_n-λ_{n-1}\right) \mathbf{u}_{n-1} \mathbf{v}_{n-1}^T$). SpecGap achieves state-of-the-art performance across multiple benchmark datasets. We present extensive ablation studies and comprehensive theoretical analyses to support our empirical results. As a parameter-free post-hoc method, SpecGap can be easily integrated into existing graph neural network models without requiring any additional training or model modification.

Jiawei Gu, Ziyue Qiao, Zechao Li

Out-of-Distribution (OOD) detection is critical for safely deploying deep models in open-world environments, where inputs may lie outside the training distribution. During inference on a model trained exclusively with In-Distribution (ID) data, we observe a salient gradient phenomenon: around an ID sample, the local gradient directions for "enhancing" that sample's predicted class remain relatively consistent, whereas OOD samples--unseen in training--exhibit disorganized or conflicting gradient directions in the same neighborhood. Motivated by this observation, we propose an inference-stage technique to short-circuit those feature coordinates that spurious gradients exploit to inflate OOD confidence, while leaving ID classification largely intact. To circumvent the expense of recomputing the logits after this gradient short-circuit, we further introduce a local first-order approximation that accurately captures the post-modification outputs without a second forward pass. Experiments on standard OOD benchmarks show our approach yields substantial improvements. Moreover, the method is lightweight and requires minimal changes to the standard inference pipeline, offering a practical path toward robust OOD detection in real-world applications.

Jiawei Gu, Ziyue Qiao, Xiao Luo

Graph Neural Networks (GNNs) have shown remarkable performance across various domains, yet they often struggle with model bias, particularly in the presence of class imbalance. This bias can lead to suboptimal performance and unfair predictions, especially for underrepresented classes. We introduce NeuBM (Neutral Bias Mitigation), a novel approach to mitigate model bias in GNNs through neutral input calibration. NeuBM leverages a dynamically updated neutral graph to estimate and correct the inherent biases of the model. By subtracting the logits obtained from the neutral graph from those of the input graph, NeuBM effectively recalibrates the model's predictions, reducing bias across different classes. Our method integrates seamlessly into existing GNN architectures and training procedures, requiring minimal computational overhead. Extensive experiments on multiple benchmark datasets demonstrate that NeuBM significantly improves the balanced accuracy and recall of minority classes, while maintaining strong overall performance. The effectiveness of NeuBM is particularly pronounced in scenarios with severe class imbalance and limited labeled data, where traditional methods often struggle. We provide theoretical insights into how NeuBM achieves bias mitigation, relating it to the concept of representation balancing. Our analysis reveals that NeuBM not only adjusts the final predictions but also influences the learning of balanced feature representations throughout the network.

Yidi Wang, Jiawei Gu, pei Xiaobing et al.

This paper studies a novel problem of out-of-distribution graph models merging, which aims to construct a generalized model from multiple graph models pre-trained on different domains with distribution discrepancy. This problem is challenging because of the difficulty in learning domain-invariant knowledge implicitly in model parameters and consolidating expertise from potentially heterogeneous GNN backbones. In this work, we propose a graph generation strategy that instantiates the mixture distribution of multiple domains. Then, we merge and fine-tune the pre-trained graph models via a MoE module and a masking mechanism for generalized adaptation. Our framework is architecture-agnostic and can operate without any source/target domain data. Both theoretical analysis and experimental results demonstrate the effectiveness of our approach in addressing the model generalization problem.

Jiawei Gu, Shangsong Liang

Effective decision-making in Large Language Models (LLMs) is essential for handling intricate tasks. However, existing approaches prioritize performance but often overlook the balance between effectiveness and computational cost. To address this, we first introduce the 3E Criteria to systematically assess the cost-effectiveness of search strategies, revealing that existing methods often trade significant efficiency for marginal performance gains. To improve LLM decision-making while maintaining efficiency, we propose the Speculative Reward Model (SRM), a plug-and-play framework that seamlessly integrates with existing search strategies. Specifically, SRM employs an external reward assigner to predict optimal actions, reducing reliance on LLMs' internal self-evaluation. And a speculative verification mechanism is used to prune suboptimal choices and guide the search toward more promising steps. We evaluate SRM on several complex decision-making tasks including mathematical reasoning, planning and numerical reasoning in specialized domains. Experimental results show that SRM reduces costs to 1/10 of the original search framework on average while maintaining effectiveness.