arXiv subjects

Yuyuan Liu

Publications and source records attributed to Yuyuan Liu.

At least 19 recordsLinked to original sources

arXiv2026

Evolutionary dynamics of collective decision-making with local social influence on static and dynamic networks

Collective decision-making is ubiquitous across the living world and artificial societies. Individuals often choose an option based on intrinsic values of options. However, individual decision-making is also swayed by neighbors' choices, generating local social influence. Hence, an important question arises naturally, yet remains unanswered: when such social influence is integrated into the individual evaluation process for option choices, how does it affect collective decision-making outcomes in structured populations modeled by graphs. To address this, we consider a baseline model of binary options with social influence and assume that individuals not only evaluate the intrinsic values of options, but are also influenced by their neighbors' choices. We propose a perceived utility function integrating these two aspects for individual decision-making. By means of theoretical analysis, we first derive the average frequency of an option on static weighted connected graphs and present the mathematical condition under which this option prevails in the population. We find that the introduction of social influence can amplify the advantage of a superior option or compensate for the deficiency of an inferior one. We also reveal that the average degree of network exerts a dual effect on collective decision outcomes. Furthermore, we consider our evolutionary model on dynamic networks switching among distinct graph configurations. Our theoretical analysis shows that the evolutionary outcomes depend not only on the average degree of each network configuration, but also on its expected duration. We perform computer simulations to verify our theoretical predictions on static and dynamic networks.

physics.soc-ph↗

arXiv2026

Git Context Controller: Manage the Context of LLM-based Agents like Git

Large language model (LLM) agents have demonstrated strong capabilities in long-horizon tasks by interleaving reasoning with tool use. However, as these agents scale to complex workflows such as software engineering and open-ended research, context management becomes a fundamental bottleneck: interaction histories grow unbounded, become costly to maintain, and are difficult to reuse across sessions and agents. We introduce \textbf{Git-Context-Controller (GCC)}, a structured context management framework inspired by software version control systems. GCC elevates agent context from a transient token stream to a persistent, navigable memory workspace with explicit operations -- \texttt{COMMIT}, \texttt{BRANCH}, \texttt{MERGE}, and \texttt{CONTEXT}, that enable milestone-based checkpointing, isolated exploration of alternative reasoning paths, and hierarchical retrieval of historical context. By organizing agent memory as a versioned file system, GCC allows agents to manage long-term goals, recover and transfer reasoning across sessions, and coordinate multi-trajectory problem solving in a principled manner. Empirically, agents equipped with GCC achieve state-of-the-art performance on both SWE-Bench and BrowseComp benchmarks. On SWE-Bench Verified, GCC improves task resolution by over 13\% relative to strong long-context baselines and outperforms 26 existing open and commercial systems, reaching over 80\% success rate. The project will be open-sourced for the research community. https://github.com/ImprintLab/git-context-controller

cs.SE↗

arXiv2026

Evolutionary dynamics in public goods games with general frequency-dependent returns

The public goods game serves as a significant paradigm for investigating the emergence and maintenance of cooperation in conflicting situations. In the traditional public goods game, the multiplication factor characterizing the synergy effect of common efforts is typically assumed to be constant. In real-world scenarios, however, investment returns are often dynamic and vary with the strategic composition of the interaction group. To date, the evolutionary dynamics of the public goods game with such frequency-dependent returns have remained not fully understood. In this work, we introduce a general frequency-dependent multiplication factor that depends on the strategy composition within the game group. Through theoretical analysis, we derive the mathematical conditions under which cooperation is favored. Our results show that whether cooperation has an evolutionary advantage over defection depends on the investment return rate in the full-contribution state of the game group, irrespective of the return rates in other states. An increase in this rate leads to a higher abundance of cooperators. Furthermore, we introduce a general frequency-dependent multiplication factor into the public goods game with peer punishment, and systematically explore its effects on the cooperation dilemma and the second-order free-rider problem. Our results highlight that the abundance of cooperators or punishers is governed solely by the investment return values in the full-cooperation and full-punishment compositions of the group. A higher return rate in the full-cooperation state facilitates the promotion of cooperation, whereas a higher return rate in the full-punishment state favors the emergence of punishment. Our theoretical findings are verified by individual-based simulations.

q-bio.PE↗

arXiv2026

Addressing Benchmarking Gaps in Large Language Models for Health and Medicine with Dynamic Red-Teaming

Large language models (LLMs) are increasingly used to answer health-related questions and support healthcare workflows, yet evidence for their safety still relies heavily on static benchmarks that can rapidly become obsolete or be optimized against. Here we introduce a Dynamic, Automatic, and Systematic (DAS) red-teaming audit framework that continuously stress-tests LLMs for health across four safety-critical axes: robustness, privacy, bias/fairness, and hallucination/factual inaccuracies. Validated against board-certified clinicians with high concordance, a suite of adversarial agents autonomously mutates health-related test cases to uncover vulnerabilities in real time. Applying DAS to 15 proprietary and open-source LLMs revealed a profound gap between high static benchmark performance and low dynamic reliability--the "Benchmarking Gap". Despite median MedQA accuracy exceeding 80\%, 94\% of previously correct answers failed under dynamic robustness testing. This brittleness generalized to the realistic, open-ended HealthBench dataset, where top-tier models exhibited failure rates exceeding 70\% and sharp shifts in model rankings across evaluations, suggesting that high scores on established static benchmarks may reflect superficial memorization. We observed similarly high failure rates across other domains: privacy leaks were elicited in 86\% of scenarios, cognitive-bias priming altered recommendations in 81\% of fairness tests, and hallucination rates exceeded 74\% in widely used models. By converting LLM safety evaluation for health from a static checklist into a living adversarial audit, DAS provides a scalable framework for surfacing latent risks before such systems are deployed in consumer-facing health assistants, clinician-facing tools, and broader healthcare workflows. Code is available at https://github.com/JZPeterPan/DAS-Medical-Red-Teaming-Agents.

cs.LG↗

arXiv2026

From Failure to Feedback: Group Revision Unlocks Hard Cases in Object-Level Grounding

Finetuning Large Vision-Language Models with reinforcement learning has emerged as a promising approach to enhance their capability in object-level grounding. However, existing methods, mainly based on GRPO, assign rewards at the response level. Such sparse reward, often criterion-induced, leads to minimal learning signals when all candidate responses fail in challenging scenarios. In this work, we propose a group-revision optimisation paradigm that enhances learning on hard cases. It begins with a sampled initial response and generates a set of revised candidates to explore improved grounding outcomes. Inspired by reward shaping, we introduce a consolidation process that quantifies each candidate's improvement over the initial attempt and converts it into informative shaping signals. These signals are used to both refine the reward and modulate the advantage, amplifying the influence of high-quality revisions. Our method achieves consistent gains across referring and reasoning segmentation, REC, and counting benchmarks compared with prior GRPO-based models. Our code is available at https://github.com/yyliu01/GroupRevision.

cs.CV↗

arXiv2026

Segmentation, Detection and Explanation: A Unified Framework for CT Appearance Reasoning

Recent progress in deep learning has significantly advanced CT image analysis, particularly for segmentation tasks. However, these advances are largely confined to image-level pattern recognition, with most methods lacking explicit anatomical or contextual reasoning. Large vision-language models introduce linguistic context into image analysis, yet most approaches typically focus on a single task, which is insufficient for clinical workflow analysis that requires multiple fine-grained types of analysis, such as anatomy detection and segmentation. In this paper, we propose a unified autoregressive framework that integrates language-guided visual reasoning into CT interpretation. Our method introduces task-routing tokens that trigger detection and segmentation heads conditioned on the hidden states of a large vision-language model, enabling coherent generation of visual outputs (e.g., masks and bounding boxes) and textual reasonings. To progressively enhance localisation accuracy and semantic clarity, we further design a "closer-look" mechanism that allows the model to perform progressive coarse-to-fine visits to regions of interest under refined fields of view. To support model training and evaluation, we curated a new multimodal CT dataset containing pixel-wise masks, bounding boxes, spatial prompts, and structured descriptions for visual objects constructed through an AI-assisted annotation process with human verification. Experiments on public benchmarks demonstrate consistent improvements over the SoTA, achieving up to 1.0% Dice on BTCV and 1.7% Dice on MosMed+, while additionally providing appearance reasoning outputs. The code and dataset will be available.

cs.CV↗

arXiv2026

AuralSAM2: Enabling SAM2 Hear Through Pyramid Audio-Visual Feature Prompting

Segment Anything Model 2 (SAM2) exhibits strong generalisation for promptable segmentation in video clips; however, its integration with the audio modality remains underexplored. Existing approaches either convert audio into visual prompts (e.g., boxes) via foundation models, or inject adapters into the image encoder for audio-visual fusion. Yet both directions fall short in human-in-the-loop scenarios due to limited prompt accuracy and increased inference overhead. In particular, these adapter-based methods often suffer from audio prompt dilution, where the signal gradually weakens as it propagates through the network. In this work, we propose AuralSAM2, which integrates audio into SAM2 while largely preserving its promptable segmentation capability. Its core module, AuralFuser, fuses audio and visual features to generate sparse and dense prompts. Guided by audio and built upon SAM2's feature pyramid, these prompts propagate auditory cues across visual layers, reinforcing cross-modal influence. To further align modalities, we introduce an audio-guided contrastive loss that emphasises auditory relevance in dominant visual features. Our method achieves notable accuracy gains on public benchmarks with only minimal impact on the interactive efficiency of promptable segmentation. Our code is available at https://github.com/yyliu01/AuralSAM2.

cs.CV↗

arXiv2026

RealICU: Do LLM Agents Understand Long-Context ICU Data? A Benchmark Beyond Behavior Imitation

Intensive care units (ICU) generate long, dense and evolving streams of clinical information, where physicians must repeatedly reassess patient states under time pressure, underscoring a clear need for reliable AI decision support. Existing ICU benchmarks typically treat historical clinician actions as ground truth. However, these actions are made under incomplete information and limited temporal context of the underlying patient state, and may therefore be suboptimal, making it difficult to assess the true reasoning capabilities of AI systems. We introduce RealICU, a hindsight-annotated benchmark for evaluating large language models (LLMs) under realistic ICU conditions, where labels are created after senior physicians review the full patient trajectory. We formulate four physician-motivated tasks: assess Patient Status, Acute Problems, Recommended Actions, and Red Flag actions that risk unsafe outcomes. We partition each trajectory with 30-min windows and release two datasets: RealICU-Gold with 930-window annotations from 94 MIMIC-IV patients, and RealICU-Scale with 11,862 windows extended by Oracle, a physician-validated LLM hindsight labeler. Existing LLMs including memory-augmented ones performed poorly on RealICU, exposing two failure modes: a recall-safety tradeoff for clinical recommendations, and an anchoring bias to early interpretations of the patient. We further introduce ICU-Evo to study structured-memory agents that improves long-horizon reasoning but does not fully eliminate safety failures. Together, RealICU provides a clinically grounded testbed for measuring and improving AI sequential decision-support in high-stakes care. Project page: https://chengzhi-leo.github.io/RealICU-Bench/

cs.AI↗

arXiv2026

Ask Patients with Patience: Enabling LLMs for Human-Centric Medical Dialogue with Grounded Reasoning

The severe shortage of medical doctors limits access to timely and reliable healthcare, leaving millions underserved. Large language models (LLMs) offer a potential solution but struggle in real-world clinical interactions. Many LLMs are not grounded in authoritative medical guidelines and fail to transparently manage diagnostic uncertainty. Their language is often rigid and mechanical, lacking the human-like qualities essential for patient trust. To address these challenges, we propose Ask Patients with Patience (APP), a multi-turn LLM-based medical assistant designed for grounded reasoning, transparent diagnoses, and human-centric interaction. APP enhances communication by eliciting user symptoms through empathetic dialogue, significantly improving accessibility and user engagement. It also incorporates Bayesian active learning to support transparent and adaptive diagnoses. The framework is built on verified medical guidelines, ensuring clinically grounded and evidence-based reasoning. To evaluate its performance, we develop a new benchmark that simulates realistic medical conversations using patient agents driven by profiles extracted from real-world consultation cases. We compare APP against SOTA one-shot and multi-turn LLM baselines. The results show that APP improves diagnostic accuracy, reduces uncertainty, and enhances user experience. By integrating medical expertise with transparent, human-like interaction, APP bridges the gap between AI-driven medical assistance and real-world clinical practice.

cs.CL↗

arXiv2026

POUR: A Provably Optimal Method for Unlearning Representations via Neural Collapse

In computer vision, machine unlearning aims to remove the influence of specific visual concepts or training images without retraining from scratch. Studies show that existing approaches often modify the classifier while leaving internal representations intact, resulting in incomplete forgetting. In this work, we extend the notion of unlearning to the representation level, deriving a three-term interplay between forgetting efficacy, retention fidelity, and class separation. Building on Neural Collapse theory, we show that the orthogonal projection of a simplex Equiangular Tight Frame (ETF) remains an ETF in a lower dimensional space, yielding a provably optimal forgetting operator. We further introduce the Representation Unlearning Score (RUS) to quantify representation-level forgetting and retention fidelity. Building on this, we introduce POUR (Provably Optimal Unlearning of Representations), a geometric projection method with closed-form (POUR-P) and a feature-level unlearning variant under a distillation scheme (POUR-D). Experiments on CIFAR-10/100 and PathMNIST demonstrate that POUR achieves effective unlearning while preserving retained knowledge, outperforming state-of-the-art unlearning methods on both classification-level and representation-level metrics.

cs.CV↗

arXiv2026

Neural Collapse-Inspired Multi-Label Federated Learning under Label-Distribution Skew

Federated Learning (FL) enables collaborative model training across distributed clients while preserving data privacy, but remains challenging when client data are highly heterogeneous. These challenges are further amplified in multi-label scenarios, where inter-label dependencies and mismatches between local and global label relationships introduce additional optimization conflicts. While most FL studies focus on single-label classification, many real-world applications are inherently multi-label and often exhibit severe label skew across clients. To address this important yet underexplored problem, we propose FedNCA-ML, a novel FL framework that aligns client representations and learns discriminative, well-clustered features inspired by Neural Collapse (NC) theory. NC describes an ideal latent geometry where each class's features collapse to their mean, forming a maximally separated simplex. FedNCA-ML further introduces an attention-based module to extract class-specific representations, enabling more balanced learning under heavy label imbalance. These class-wise representations are then aligned via a shared NC-inspired structure, mitigating inter-client conflicts induced by heterogeneous local data and inconsistent label dependencies. In addition, we design regularisation losses to encourage compact and consistent feature clustering in the latent space. Experiments on five benchmark datasets under nine FL settings demonstrate the effectiveness of the proposed method, achieving improvements of up to 3.92% in class-wise AUC and 4.93% in class-wise F1 score.

cs.CV↗

arXiv2026

Coevolutionary dynamics of cooperation, risk, and cost in collective risk games

Addressing both natural and societal challenges requires collective cooperation. Studies on collective-risk social dilemmas have shown that individual decisions are influenced by the perceived risk of collective failure. However, existing feedback evolving game models often focus on a single feedback mechanism, such as the coupling between cooperation and risk or between cooperation and cost. In many real-world scenarios, however, the level of cooperation, the cost of cooperating, and the collective risk are dynamically interlinked. Here, we present an evolutionary game model that considers the interplay of these three variables. Our analysis shows that the worst-case scenario, characterized by full defection, maximum risk, and the highest cost of cooperation, remains a stable evolutionary attractor. Nevertheless, cooperation can emerge and persist because the system also supports stable equilibria with non-zero cooperation. The system exhibits multistability, meaning that different initial conditions lead to either sustained cooperation or a tragedy of the commons. These findings highlight that initial levels of cooperation, cost, and risk collectively determine whether a population can avert a tragic outcome.

nlin.AO↗

arXiv2026

Evo: Autoregressive-Diffusion Large Language Models with Evolving Balance

We introduce \textbf{Evo}, a duality latent trajectory model that bridges autoregressive (AR) and diffusion-based language generation within a continuous evolutionary generative framework. Rather than treating AR decoding and diffusion generation as separate paradigms, Evo reconceptualizes text generation as a latent flow: each token is associated with a vector-valued embedding that evolves over a progression variable $t_i \in [0, 1]$, indicating its semantic maturity. Low $t_i$ values correspond to confident AR-like refinement, while high values invoke diffusion-style planning, allowing the model to adaptively balance AR and diffusion based on uncertainty. Theoretically, we show that both AR and diffusion models emerge as discretizations of a shared probability flow, and we derive Evo's training objective from a unified variational ELBO. The model is implemented as a time-conditioned Transformer governed by a shared vector field, trained end-to-end to jointly infer latent codes and their progression times. During decoding, Evo performs efficient, semantics-aware refinement, achieving high-quality outputs without sacrificing speed. Empirically, Evo 8B achieves state-of-the-art or highly competitive results on 15 diverse benchmarks, including reasoning (GSM8K, ARC-C), code generation (HumanEval, MBPP), and general language understanding, while maintaining fast inference speed. Our results demonstrate that Evo delivers a new paradigm for LLM design with strong generation quality, robust symbolic reasoning, and decoding efficiency.

cs.LG↗

arXiv2025

Bridging Generative and Discriminative Noisy-Label Learning via Direction-Agnostic EM Formulation

Although noisy-label learning is often approached with discriminative methods for simplicity and speed, generative modeling offers a principled alternative by capturing the joint mechanism that produces features, clean labels, and corrupted observations. However, prior work typically (i) introduces extra latent variables and heavy image generators that bias training toward reconstruction, (ii) fixes a single data-generating direction ($Y\rightarrow\!X$ or $X\rightarrow\!Y$), limiting adaptability, and (iii) assumes a uniform prior over clean labels, ignoring instance-level uncertainty. We propose a single-stage, EM-style framework for generative noisy-label learning that is \emph{direction-agnostic} and avoids explicit image synthesis. First, we derive a single Expectation-Maximization (EM) objective whose E-step specializes to either causal orientation without changing the overall optimization. Second, we replace the intractable $p(X\mid Y)$ with a dataset-normalized discriminative proxy computed using a discriminative classifier on the finite training set, retaining the structural benefits of generative modeling at much lower cost. Third, we introduce \emph{Partial-Label Supervision} (PLS), an instance-specific prior over clean labels that balances coverage and uncertainty, improving data-dependent regularization. Across standard vision and natural language processing (NLP) noisy-label benchmarks, our method achieves state-of-the-art accuracy, lower transition-matrix estimation error, and substantially less training compute than current generative and discriminative baselines. Code: https://github.com/lfb-1/GNL

cs.CV↗

arXiv2025

MedCutMix: A Data-Centric Approach to Improve Radiology Vision-Language Pre-training with Disease Awareness

Vision-Language Pre-training (VLP) is drawing increasing interest for its ability to minimize manual annotation requirements while enhancing semantic understanding in downstream tasks. However, its reliance on image-text datasets poses challenges due to privacy concerns and the high cost of obtaining paired annotations. Data augmentation emerges as a viable strategy to address this issue, yet existing methods often fall short of capturing the subtle and complex variations in medical data due to limited diversity. To this end, we propose MedCutMix, a novel multi-modal disease-centric data augmentation method. MedCutMix performs diagnostic sentence CutMix within medical reports and establishes the cross-attention between the diagnostic sentence and medical image to guide attentive manifold mix within the imaging modality. Our approach surpasses previous methods across four downstream radiology diagnosis datasets, highlighting its effectiveness in enhancing performance and generalizability in radiology VLP.

cs.CV↗

arXiv2025

Meta-Learned Modality-Weighted Knowledge Distillation for Robust Multi-Modal Learning with Missing Data

In multi-modal learning, some modalities are more influential than others, and their absence can have a significant impact on classification/segmentation accuracy. Addressing this challenge, we propose a novel approach called Meta-learned Modality-weighted Knowledge Distillation (MetaKD), which enables multi-modal models to maintain high accuracy even when key modalities are missing. MetaKD adaptively estimates the importance weight of each modality through a meta-learning process. These learned importance weights guide a pairwise modality-weighted knowledge distillation process, allowing high-importance modalities to transfer knowledge to lower-importance ones, resulting in robust performance despite missing inputs. Unlike previous methods in the field, which are often task-specific and require significant modifications, our approach is designed to work in multiple tasks (e.g., segmentation and classification) with minimal adaptation. Experimental results on five prevalent datasets, including three Brain Tumor Segmentation datasets (BraTS2018, BraTS2019 and BraTS2020), the Alzheimer's Disease Neuroimaging Initiative (ADNI) classification dataset and the Audiovision-MNIST classification dataset, demonstrate the proposed model is able to outperform the compared models by a large margin. The code is available at https://github.com/billhhh/MetaKD.

cs.CV↗

arXiv2025

Strategic competition in informal risk sharing mechanism versus collective index insurance

The frequent occurrence of natural disasters has posed significant challenges to society, necessitating the urgent development of effective risk management strategies. From the early informal community-based risk sharing mechanisms to modern formal index insurance products, risk management tools have continuously evolved. Although index insurance provides an effective risk transfer mechanism in theory, it still faces the problems of basis risk and pricing in practice. At the same time, in the presence of informal community risk sharing mechanisms, the competitiveness of index insurance deserves further investigation. Here we propose a three-strategy evolutionary game model, which simultaneously examines the competitive relationship between formal index insurance purchasing (I), informal risk sharing strategies (S), and complete non-insurance (A). Furthermore, we introduce a method for calculating insurance company profits to aid in the optimal pricing of index insurance products. We find that basis risk and risk loss ratio have significant impacts on insurance adoption rate. Under scenarios with low basis risk and high loss ratios, index insurance is more popular; meanwhile, when the loss ratio is moderate, an informal risk sharing strategy is the preferred option. Conversely, when the loss ratio is low, individuals tend to forego any insurance. Furthermore, accurately assessing the degree of risk aversion and determining the appropriate ratio of risk sharing are crucial for predicting the future market sales of index insurance.

q-fin.RM↗

arXiv2025

Agentic Reasoning: A Streamlined Framework for Enhancing LLM Reasoning with Agentic Tools

We introduce Agentic Reasoning, a framework that enhances large language model (LLM) reasoning by integrating external tool-using agents. Agentic Reasoning dynamically leverages web search, code execution, and structured memory to address complex problems requiring deep research. A key innovation in our framework is the Mind-Map agent, which constructs a structured knowledge graph to store reasoning context and track logical relationships, ensuring coherence in long reasoning chains with extensive tool usage. Additionally, we conduct a comprehensive exploration of the Web-Search agent, leading to a highly effective search mechanism that surpasses all prior approaches. When deployed on DeepSeek-R1, our method achieves a new state-of-the-art (SOTA) among public models and delivers performance comparable to OpenAI Deep Research, the leading proprietary model in this domain. Extensive ablation studies validate the optimal selection of agentic tools and confirm the effectiveness of our Mind-Map and Web-Search agents in enhancing LLM reasoning. The code is at: https://github.com/theworldofagents/Agentic-Reasoning

cs.AI↗