arXiv subjects

Mengdi Wang

Publications and source records attributed to Mengdi Wang.

At least 91 records · Page 5Linked to original sources

arXiv2025

Zoom in, Click out: Unlocking and Evaluating the Potential of Zooming for GUI Grounding

Grounding is a fundamental capability for building graphical user interface (GUI) agents. Although existing approaches rely on large-scale bounding box supervision, they still face various challenges, such as cross-platform generalization, complex layout analysis, and fine-grained element localization. In this paper, we investigate zoom as a strong yet underexplored prior for GUI grounding, and propose a training-free method, ZoomClick. By characterizing four key properties of zoom (i.e., pre-zoom, depth, shrink size, minimal crop size), we unlock its full capabilities for dynamic spatial focusing and adaptive context switching. Experiments demonstrate that our method significantly boosts the performance of both general vision-language and specialized GUI grounding models, achieving state-of-the-art results on several mainstream benchmarks; for example, UI-Venus-72B attains a 73.1% success rate on ScreenSpot-Pro. Furthermore, we present GUIZoom-Bench, a benchmark for evaluating model adaptability to zoom, aiming to inspire future research on improving zoom for further training and test-time scaling in GUI grounding tasks.

cs.CV↗

arXiv2025

CycleSL: Server-Client Cyclical Update Driven Scalable Split Learning

Split learning emerges as a promising paradigm for collaborative distributed model training, akin to federated learning, by partitioning neural networks between clients and a server without raw data exchange. However, sequential split learning suffers from poor scalability, while parallel variants like parallel split learning and split federated learning often incur high server resource overhead due to model duplication and aggregation, and generally exhibit reduced model performance and convergence owing to factors like client drift and lag. To address these limitations, we introduce CycleSL, a novel aggregation-free split learning framework that enhances scalability and performance and can be seamlessly integrated with existing methods. Inspired by alternating block coordinate descent, CycleSL treats server-side training as an independent higher-level machine learning task, resampling client-extracted features (smashed data) to mitigate heterogeneity and drift. It then performs cyclical updates, namely optimizing the server model first, followed by client updates using the updated server for gradient computation. We integrate CycleSL into previous algorithms and benchmark them on five publicly available datasets with non-iid data distribution and partial client attendance. Our empirical findings highlight the effectiveness of CycleSL in enhancing model performance. Our source code is available at https://gitlab.lrz.de/hctl/CycleSL.

cs.LG↗

arXiv2025

MMaDA-Parallel: Multimodal Large Diffusion Language Models for Thinking-Aware Editing and Generation

While thinking-aware generation aims to improve performance on complex tasks, we identify a critical failure mode where existing sequential, autoregressive approaches can paradoxically degrade performance due to error propagation. To systematically analyze this issue, we propose ParaBench, a new benchmark designed to evaluate both text and image output modalities. Our analysis using ParaBench reveals that this performance degradation is strongly correlated with poor alignment between the generated reasoning and the final image. To resolve this, we propose a parallel multimodal diffusion framework, MMaDA-Parallel, that enables continuous, bidirectional interaction between text and images throughout the entire denoising trajectory. MMaDA-Parallel is trained with supervised finetuning and then further optimized by Parallel Reinforcement Learning (ParaRL), a novel strategy that applies semantic rewards along the trajectory to enforce cross-modal consistency. Experiments validate that our model significantly improves cross-modal alignment and semantic consistency, achieving a 6.9\% improvement in Output Alignment on ParaBench compared to the state-of-the-art model, Bagel, establishing a more robust paradigm for thinking-aware image synthesis. Our code is open-sourced at https://github.com/tyfeld/MMaDA-Parallel

cs.CV↗

arXiv2025

Quantitative bounds in a popular polynomial Szemerédi theorem

We obtain polylogarithmic bounds in the polynomial Szemerédi theorem when the polynomials have distinct degrees and zero constant terms. Specifically, let $P_1, \dots, P_m \in \mathbb Z[y]$ be polynomials with distinct degrees, each having zero constant term. Then there exists a constant $c = c(P_1,\dots,P_m) > 0$ such that any subset $A \subset \{1,2,\dots,N\}$ of density at least $(\log N)^{-c}$ contains a nontrivial polynomial progression of the form $x, x+P_1(y), \dots, x+P_m(y)$. In addition, we prove an effective ``popular'' version, showing that every dense subset $A$ has some non-zero $y$ such that the number of polynomial progressions in $A$ with this difference $y$ is asymptotically at least as large as in a random set of the same density as $A$.

math.NT↗

arXiv2025

Generative AI for Biosciences: Emerging Threats and Roadmap to Biosecurity

The rapid adoption of generative artificial intelligence (GenAI) in the biosciences is transforming biotechnology, medicine, and synthetic biology. Yet this advancement is intrinsically linked to new vulnerabilities, as GenAI lowers the barrier to misuse and introduces novel biosecurity threats, such as generating synthetic viral proteins or toxins. These dual-use risks are often overlooked, as existing safety guardrails remain fragile and can be circumvented through deceptive prompts or jailbreak techniques. In this Perspective, we first outline the current state of GenAI in the biosciences and emerging threat vectors ranging from jailbreak attacks and privacy risks to the dual-use challenges posed by autonomous AI agents. We then examine urgent gaps in regulation and oversight, drawing on insights from 130 expert interviews across academia, government, industry, and policy. A large majority ($\approx 76$\%) expressed concern over AI misuse in biology, and 74\% called for the development of new governance frameworks. Finally, we explore technical pathways to mitigation, advocating a multi-layered approach to GenAI safety. These defenses include rigorous data filtering, alignment with ethical principles during development, and real-time monitoring to block harmful requests. Together, these strategies provide a blueprint for embedding security throughout the GenAI lifecycle. As GenAI becomes integrated into the biosciences, safeguarding this frontier requires an immediate commitment to both adaptive governance and secure-by-design technologies.

cs.CR↗

arXiv2025

Alita-G: Self-Evolving Generative Agent for Agent Generation

Large language models (LLMs) have been shown to perform better when scaffolded into agents with memory, tools, and feedback. Beyond this, self-evolving agents have emerged, but current work largely limits adaptation to prompt rewriting or failure retries. Therefore, we present ALITA-G, a self-evolution framework that transforms a general-purpose agent into a domain expert by systematically generating, abstracting, and curating Model Context Protocol (MCP) tools. In this framework, a generalist agent executes a curated suite of target-domain tasks and synthesizes candidate MCPs from successful trajectories. These are then abstracted to parameterized primitives and consolidated into an MCP Box. At inference time, ALITA-G performs retrieval-augmented MCP selection with the help of each tool's descriptions and use cases, before executing an agent equipped with the MCP Executor. Across several benchmarks GAIA, PathVQA, and Humanity's Last Exam, ALITA-G attains strong gains while reducing computation costs. On GAIA validation, it achieves 83.03% pass@1 and 89.09% pass@3, establishing a new state-of-the-art result while reducing mean tokens per example by approximately 15% relative to a strong baseline agent. ALITA-G thus provides a principled pathway from generalist capability to reusable, domain-specific competence, improving both accuracy and efficiency on complex reasoning tasks.

cs.AI↗

arXiv2025

Does Thinking More always Help? Mirage of Test-Time Scaling in Reasoning Models

Recent trends in test-time scaling for reasoning models (e.g., OpenAI o1, DeepSeek R1) have led to a popular belief that extending thinking traces using prompts like "Wait" or "Let me rethink" can improve performance. This raises a natural question: Does thinking more at test-time truly lead to better reasoning? To answer this question, we perform a detailed empirical study across models and benchmarks, which reveals a consistent pattern of initial performance improvements from additional thinking followed by a decline, due to "overthinking". To understand this non-monotonic trend, we consider a simple probabilistic model, which reveals that additional thinking increases output variance-creating an illusion of improved reasoning while ultimately undermining precision. Thus, observed gains from "more thinking" are not true indicators of improved reasoning, but artifacts stemming from the connection between model uncertainty and evaluation metric. This suggests that test-time scaling through extended thinking is not an effective way to utilize the inference thinking budget. Recognizing these limitations, we introduce an alternative test-time scaling approach, parallel thinking, inspired by Best-of-N sampling. Our method generates multiple independent reasoning paths within the same inference budget and selects the most consistent response via majority vote, achieving up to 20% higher accuracy compared to extended thinking. This provides a simple yet effective mechanism for test-time scaling of reasoning models.

cs.AI↗

arXiv2025

On the Role of Preference Variance in Preference Optimization

Direct Preference Optimization (DPO) has emerged as an important approach for learning from human preferences in aligning large language models (LLMs). However, collecting human preference data is costly and inefficient, motivating methods to reduce the required annotations. In this work, we investigate the impact of \emph{preference variance} (PVar), which measures the variance in model preferences when comparing pairs of responses, on the effectiveness of DPO training. We provide a theoretical insight by establishing an upper bound on the DPO gradient norm for any given prompt, showing it is controlled by the PVar of that prompt. This implies that prompts with low PVar can only produce small gradient updates, making them less valuable for learning. We validate this finding by fine-tuning LLMs with preferences generated by a reward model, evaluating on two benchmarks (AlpacaEval 2.0 and Arena-Hard). Experimental results demonstrate that prompts with higher PVar outperform randomly selected prompts or those with lower PVar. We also show that our PVar-based selection method is robust, when using smaller reward models (1B, 3B) for selection. Notably, in a separate experiment using the original human annotations from the UltraFeedback dataset, we found that training on only the top 10\% of prompts with the highest PVar yields better evaluation performance than training on the full dataset, highlighting the importance of preference variance in identifying informative examples for efficient LLM alignment.

cs.CL↗

arXiv2025

ChipGPT: How far are we from natural language hardware design

As large language models (LLMs) like ChatGPT exhibited unprecedented machine intelligence, it also shows great performance in assisting hardware engineers to realize higher-efficiency logic design via natural language interaction. To estimate the potential of the hardware design process assisted by LLMs, this work attempts to demonstrate an automated design environment that explores LLMs to generate hardware logic designs from natural language specifications. To realize a more accessible and efficient chip development flow, we present a scalable four-stage zero-code logic design framework based on LLMs without retraining or finetuning. At first, the demo, ChipGPT, begins by generating prompts for the LLM, which then produces initial Verilog programs. Second, an output manager corrects and optimizes these programs before collecting them into the final design space. Eventually, ChipGPT will search through this space to select the optimal design under the target metrics. The evaluation sheds some light on whether LLMs can generate correct and complete hardware logic designs described by natural language for some specifications. It is shown that ChipGPT improves programmability, and controllability, and shows broader design optimization space compared to prior work and native LLMs alone.

cs.AI↗

arXiv2025

Demystifying Reinforcement Learning in Agentic Reasoning

Recently, the emergence of agentic RL has showcased that RL could also effectively improve the agentic reasoning ability of LLMs, yet the key design principles and optimal practices remain unclear. In this work, we conduct a comprehensive and systematic investigation to demystify reinforcement learning in agentic reasoning from three key perspectives: data, algorithm, and reasoning mode. We highlight our key insights: (i) Replacing stitched synthetic trajectories with real end-to-end tool-use trajectories yields a far stronger SFT initialization; high-diversity, model-aware datasets sustain exploration and markedly improve RL performance. (ii) Exploration-friendly techniques are crucial for agentic RL, such as clip higher, overlong reward shaping, and maintaining adequate policy entropy could improve the training efficiency. (iii) A deliberative strategy with fewer tool calls outperforms frequent tool calls or verbose self-reasoning, improving tool efficiency and final accuracy. Together, these simple practices consistently enhance agentic reasoning and training efficiency, achieving strong results on challenging benchmarks with smaller models, and establishing a practical baseline for future agentic RL research. Beyond these empirical insights, we further contribute a high-quality, real end-to-end agentic SFT dataset along with a high-quality RL dataset, and demonstrate the effectiveness of our insights in boosting the agentic reasoning ability of LLMs across four challenging benchmarks, including AIME2024/AIME2025, GPQA-Diamond, and LiveCodeBench-v6. With our recipes, 4B-sized models could also achieve superior agentic reasoning performance compared to 32B-sized models. Code and models: https://github.com/Gen-Verse/Open-AgentRL

cs.CL↗

arXiv2025

SafeProtein: Red-Teaming Framework and Benchmark for Protein Foundation Models

Proteins play crucial roles in almost all biological processes. The advancement of deep learning has greatly accelerated the development of protein foundation models, leading to significant successes in protein understanding and design. However, the lack of systematic red-teaming for these models has raised serious concerns about their potential misuse, such as generating proteins with biological safety risks. This paper introduces SafeProtein, the first red-teaming framework designed for protein foundation models to the best of our knowledge. SafeProtein combines multimodal prompt engineering and heuristic beam search to systematically design red-teaming methods and conduct tests on protein foundation models. We also curated SafeProtein-Bench, which includes a manually constructed red-teaming benchmark dataset and a comprehensive evaluation protocol. SafeProtein achieved continuous jailbreaks on state-of-the-art protein foundation models (up to 70% attack success rate for ESM3), revealing potential biological safety risks in current protein foundation models and providing insights for the development of robust security protection technologies for frontier models. The codes will be made publicly available at https://github.com/jigang-fan/SafeProtein.

cs.LG↗

arXiv2025

Binding Affinity Prediction: From Conventional to Machine Learning-Based Approaches

Protein-ligand binding is the process by which a small molecule (drug or inhibitor) attaches to a target protein. Binding affinity, which characterizes the strength of biomolecular interactions, is essential for tackling diverse challenges in life sciences, including therapeutic design, protein engineering, enzyme optimization, and elucidating biological mechanisms. Much work has been devoted to predicting binding affinity over the past decades. Here, we review recent significant works, with a focus on methods, evaluation strategies, and benchmark datasets. We note growing use of both traditional machine learning and deep learning models for predicting binding affinity, accompanied by an increasing amount of data on proteins and small drug-like molecules. With improved predictive performance and the FDA's phasing out of animal testing, AI-driven in silico models, such as AI virtual cells (AIVCs), are poised to advance binding affinity prediction; reciprocally, progress in building binding affinity predictors can refine AIVCs. Future efforts in binding affinity prediction and AI-driven in silico models can enhance the simulation of temporal dynamics, cell-type specificity, and multi-omics integration to support more accurate and personalized outcomes.

q-bio.QM↗

arXiv2025

DISC: Dynamic Decomposition Improves LLM Inference Scaling

Inference scaling methods for LLMs often rely on decomposing problems into steps (or groups of tokens), followed by sampling and selecting the best next steps. However, these steps and their sizes are often predetermined or manually designed based on domain knowledge. We propose dynamic decomposition, a method that adaptively and automatically partitions solution and reasoning traces into manageable steps during inference. By more effectively allocating compute -- particularly through subdividing challenging steps and prioritizing their sampling -- dynamic decomposition significantly improves inference efficiency. Experiments on benchmarks such as APPS, MATH, and LiveCodeBench demonstrate that dynamic decomposition outperforms static approaches, including token-level, sentence-level, and single-step decompositions, reducing the pass@10 error rate by 5.0%, 6.7%, and 10.5% respectively. These findings highlight the potential of dynamic decomposition to improve a wide range of inference scaling techniques.

cs.LG↗

arXiv2025

Detecting Post-generation Edits to Watermarked LLM Outputs via Combinatorial Watermarking

Watermarking has become a key technique for proprietary language models, enabling the distinction between AI-generated and human-written text. However, in many real-world scenarios, LLM-generated content may undergo post-generation edits, such as human revisions or even spoofing attacks, making it critical to detect and localize such modifications. In this work, we introduce a new task: detecting post-generation edits locally made to watermarked LLM outputs. To this end, we propose a combinatorial pattern-based watermarking framework, which partitions the vocabulary into disjoint subsets and embeds the watermark by enforcing a deterministic combinatorial pattern over these subsets during generation. We accompany the combinatorial watermark with a global statistic that can be used to detect the watermark. Furthermore, we design lightweight local statistics to flag and localize potential edits. We introduce two task-specific evaluation metrics, Type-I error rate and detection accuracy, and evaluate our method on open-source LLMs across a variety of editing scenarios, demonstrating strong empirical performance in edit localization.

cs.LG↗

arXiv2025

MMaDA: Multimodal Large Diffusion Language Models

We introduce MMaDA, a novel class of multimodal diffusion foundation models designed to achieve superior performance across diverse domains such as textual reasoning, multimodal understanding, and text-to-image generation. The approach is distinguished by three key innovations: (i) MMaDA adopts a unified diffusion architecture with a shared probabilistic formulation and a modality-agnostic design, eliminating the need for modality-specific components. This architecture ensures seamless integration and processing across different data types. (ii) We implement a mixed long chain-of-thought (CoT) fine-tuning strategy that curates a unified CoT format across modalities. By aligning reasoning processes between textual and visual domains, this strategy facilitates cold-start training for the final reinforcement learning (RL) stage, thereby enhancing the model's ability to handle complex tasks from the outset. (iii) We propose UniGRPO, a unified policy-gradient-based RL algorithm specifically tailored for diffusion foundation models. Utilizing diversified reward modeling, UniGRPO unifies post-training across both reasoning and generation tasks, ensuring consistent performance improvements. Experimental results demonstrate that MMaDA-8B exhibits strong generalization capabilities as a unified multimodal foundation model. It surpasses powerful models like LLaMA-3-7B and Qwen2-7B in textual reasoning, outperforms Show-o and SEED-X in multimodal understanding, and excels over SDXL and Janus in text-to-image generation. These achievements highlight MMaDA's effectiveness in bridging the gap between pretraining and post-training within unified diffusion architectures, providing a comprehensive framework for future research and development. We open-source our code and trained models at: https://github.com/Gen-Verse/MMaDA

cs.CV↗

arXiv2025

Co-Evolving LLM Coder and Unit Tester via Reinforcement Learning

We propose CURE, a novel reinforcement learning framework with a dedicated reward design that co-evolves coding and unit test generation capabilities based on their interaction outcomes, without any ground-truth code as supervision. This approach enables flexible and scalable training and allows the unit tester to learn directly from the coder's mistakes. Our derived ReasonFlux-Coder-7B and 14B models improve code generation accuracy by 5.3% and Best-of-N accuracy by 9.0% after optimization on Qwen2.5-Instruct models, outperforming similarly sized Qwen-Coder, DeepSeek-Coder, and Seed-Coder. They naturally extend to downstream tasks such as test-time scaling and agentic coding-achieving a 8.1% improvement over the base model. For the long-CoT model, our ReasonFlux-Coder-4B consistently outperforms Qwen3-4B while achieving 64.8% inference efficiency in unit test generation. Notably, we also find that our model can serve as an effective reward model for reinforcement learning on base models. Project: https://github.com/Gen-Verse/CURE

cs.CL↗

arXiv2025

ReasonFlux-PRM: Trajectory-Aware PRMs for Long Chain-of-Thought Reasoning in LLMs

Process Reward Models (PRMs) have recently emerged as a powerful framework for supervising intermediate reasoning steps in large language models (LLMs). Previous PRMs are primarily trained on model final output responses and struggle to evaluate intermediate thinking trajectories robustly, especially in the emerging setting of trajectory-response outputs generated by frontier reasoning models like Deepseek-R1. In this work, we introduce ReasonFlux-PRM, a novel trajectory-aware PRM explicitly designed to evaluate the trajectory-response type of reasoning traces. ReasonFlux-PRM incorporates both step-level and trajectory-level supervision, enabling fine-grained reward assignment aligned with structured chain-of-thought data. We adapt ReasonFlux-PRM to support reward supervision under both offline and online settings, including (i) selecting high-quality model distillation data for downstream supervised fine-tuning of smaller models, (ii) providing dense process-level rewards for policy optimization during reinforcement learning, and (iii) enabling reward-guided Best-of-N test-time scaling. Empirical results on challenging downstream benchmarks such as AIME, MATH500, and GPQA-Diamond demonstrate that ReasonFlux-PRM-7B selects higher quality data than strong PRMs (e.g., Qwen2.5-Math-PRM-72B) and human-curated baselines. Furthermore, our derived ReasonFlux-PRM-7B yields consistent performance improvements, achieving average gains of 12.1% in supervised fine-tuning, 4.5% in reinforcement learning, and 6.3% in test-time scaling. We also release our efficient ReasonFlux-PRM-1.5B for resource-constrained applications and edge deployment. Project: https://github.com/Gen-Verse/ReasonFlux

cs.CL↗

arXiv2025

FragmentGPT: A Unified GPT Model for Fragment Growing, Linking, and Merging in Molecular Design

Fragment-Based Drug Discovery (FBDD) is a popular approach in early drug development, but designing effective linkers to combine disconnected molecular fragments into chemically and pharmacologically viable candidates remains challenging. Further complexity arises when fragments contain structural redundancies, like duplicate rings, which cannot be addressed by simply adding or removing atoms or bonds. To address these challenges in a unified framework, we introduce FragmentGPT, which integrates two core components: (1) a novel chemically-aware, energy-based bond cleavage pre-training strategy that equips the GPT-based model with fragment growing, linking, and merging capabilities, and (2) a novel Reward Ranked Alignment with Expert Exploration (RAE) algorithm that combines expert imitation learning for diversity enhancement, data selection and augmentation for Pareto and composite score optimality, and Supervised Fine-Tuning (SFT) to align the learner policy with multi-objective goals. Conditioned on fragment pairs, FragmentGPT generates linkers that connect diverse molecular subunits while simultaneously optimizing for multiple pharmaceutical goals. It also learns to resolve structural redundancies-such as duplicated fragments-through intelligent merging, enabling the synthesis of optimized molecules. FragmentGPT facilitates controlled, goal-driven molecular assembly. Experiments and ablation studies on real-world cancer datasets demonstrate its ability to generate chemically valid, high-quality molecules tailored for downstream drug discovery tasks.

cs.LG↗