As large language models (LLMs) increasingly serve as autonomous agents in social simulations, ensuring their ability to understand and comply with legal norms is essential. Yet, current LLM agents frequently exhibit reward hacking (RH) behaviors by optimizing metrics at the expense of norm adherence, undermining simulation fidelity and limiting deployment. We introduce a TBC-TBA self-refine learning multi-agent framework that enables dynamic normative adaptation through iterative multi-agent feedback. This framework integrates Think-Before-Chat (social feedback processing) and Think-Before-Act (norm-guided decision making) phases, allowing agents to progressively refine their normative understanding via structured interaction cycles. Across five mainstream LLMs and 100 legal scenarios, we found that while LLMs partially recognize legal norms, they systematically exhibit RH behaviors with illegal action rates (IAR) of 14.29–37.11%. Comparison with human cognition reveals alignment in moral reasoning but sharp divergence in risk perception and probability distortion. To address these deficits, we introduce four methods to improve LLM's normative compliance. Dynamic Norm Learning Mechanism (DNLM) serves as the core, using a psychologically grounded identify–infer–implement process that reduces IAR by 15.78% on average and delivers the most significant improvement. We also introduce Deep MaxPain (DMP) for consequence based deterrence, Norm Analysis Chain-of-Thought (NA-CoT) for structured reasoning, and Few-shot Norm Learning (FNL) for case based acquisition, all of them enhance compliance. Our findings show that LLM agents can better follow legal norms when equipped with structured self-refine learning and psychologically informed mechanisms. This work improves social alignment in multi-agent systems and opens avenues for future research on scalable, norm-compliant autonomous agents.The code and data are publicly available on GitHub.
论文检索
输入标题、作者或关键词，从 990 篇学术成果中精准定位
As generative AI makes polished prose cheap to produce, users can no longer rely on fluency as a proxy for truth. We call this failure mode the Fluency Trap: users trust fluent hallucinations while also discounting accurate content once it is disclosed as AI-generated. Binary "Made with AI" labels respond with authorship disclosure, but they do not show what supports a claim. We propose Provenance Density, an evidence-visualization interface that shows the density of verified claims in a text. In a user study with 81 participants, an idealized Provenance Density interface produced a large discernment gap between truth and fabrication (+4.15 points, d=1.82), whereas participants given no signal showed no detectable discrimination. A technical audit with 200 samples shows that retrieval density alone is insufficient; unexpectedly, the Consistency Veto carries most of the discriminative signal on dynamic queries. As AI-generated content becomes indistinguishable from human writing, effective transparency must move from authorship disclosure toward evidence visualization.
Large Language Models (LLMs) can generate Computer-Aided Design (CAD), yet lack physical comprehension required for reliable engineering design. Instead of attempting to implicitly learn physical laws from data, we propose a Hybrid Agentic-Physical Architecture that embeds validated knowledge-based engineering tools directly into the decision-making loop of autonomous AI agents. In this framework, engineering design is formulated as a closed-loop, sequential decision-making process guided by explicit physical verification. Based on a load case, dedicated agents iteratively plan, generate, evaluate, and revise engineering designs using knowledge-based tools as a feedback signal. We introduce a benchmark dataset and metrics for assessing functional validity in generative CAD. Our system generates more complex and physically verified designs, with a 4.2x increase in structural complexity and improving compile rate by 3.5% compared to similar agentic methods. The codebase, prompts and dataset will be made publicly available to support reproducibility and future research.
Fragmented image recovery is of significant importance in computer vision, such as cultural relic and artwork restoration, archival document recovery, and digital forensics. The goal is to recover the original image topology from an unordered set of fragments and spatially align and stitch them together. The adjacency relationships among fragments are discrete, sparse, and highly structured, making it difficult for traditional methods to effectively handle global topological consistency. To address this challenge, we propose a fragment adjacency recovery method based on a conditional graph diffusion model. First, we perform discrete denoising pretraining with structural masking to learn structure-aware node representations from perturbed adjacency matrices, using graph neural networks for message passing. Building on this, we design a masked discrete diffusion process tailored for fragment reconstruction, which progressively restores the connectivity between fragments. Furthermore, to enhance the controllability of the generation process, we introduce a topology-guided mechanism that steers the generation of adjacency structures via a topological scoring function, ensuring that the reconstructed fragment graph satisfies global topological constraints. Experimental results demonstrate that our method achieves state-of-the-art performance on hand-torn calligraphy, painting replica datasets and document datasets, outperforming existing approaches in both accuracy and robustness.
The Metaverse is envisioned to support immersive, large-scale social interactions via virtual reality (VR) displays. However, scalability remains a major bottleneck: as the number of concurrent users grows, tracking and updating display content incurs quadratic overhead, often limiting a shared virtual space to only a few dozen participants. Our key observation is that user attention in social VR is highly selective, with users primarily focusing on socially relevant peers rather than all visible users. Motivated by this observation, we propose SAGE, a social-aware graph-based VR display framework that enables personalized displays based on inferred social relevance. SAGE introduces a dual-graph learning architecture to jointly model long-term social structures and short-term spatiotemporal co-presence patterns, generating complementary interest scores for display prioritization. Based on these scores, we formulate scalable VR display support as a multi-dimensional resource allocation problem and design a lightweight coordination mechanism with provable guarantees, including incentive compatibility and individual rationality. Experiments on Metaverse datasets show that SAGE improves interaction-relevance prediction by 11.64% and increases social welfare by up to 2.4× compared to state-of-the-art schemes. It scales to support up to 1,000 concurrent users and remains robust against strategic manipulation.
Autonomous AI agents on embedded platforms require real-time, risk-aware scheduling under resource and thermal constraints. Classical heuristics struggle with workload irregularity, tabular regressors discard structural information, and model-free reinforcement learning (RL) risks overheating. We introduce GraphPerf-RT, an AI technology achieving deep learning accuracy at heuristic speeds (2-7ms). GraphPerf-RT is, to our knowledge, the first graph-grounded infrastructure unifying task DAG topology, CFG-derived code semantics, and runtime context (per-core DVFS, thermal state, utilization) in a heterogeneous graph with typed edges encoding precedence, placement, and contention. The architecture supports multi-task evidential heads with Normal-Inverse-Gamma uncertainty; we validate on makespan prediction for risk-aware scheduling. Experiments on three ARM platforms (Jetson TX2, Orin NX, RUBIK Pi) achieve R^2 = 0.81 on log-transformed makespan with Spearman rho = 0.95 and conservative uncertainty calibration (PICP = 99.9% at 95% confidence). Integration with four RL methods demonstrates that multi-agent model-based RL with GraphPerf-RT as the world model achieves 66% makespan reduction and 82% energy reduction versus model-free baselines, with zero thermal violations.
Modern generative models for limit order books (LOBs) can reproduce realistic market dynamics, but they remain fundamentally passive: they either model what typically happens without accounting for hypothetical future market conditions, or they require interaction with another agent to explore alternative outcomes. This limits their usefulness for stress testing, scenario analysis, and decision-making. We propose DiffLOB, a regime-conditioned Diffusion model for controllable and counterfactual generation of LOB trajectories. DiffLOB explicitly conditions the generative process on future market regimes—including trend, volatility, liquidity, and order-flow imbalance, which enables the model to answer counterfactual queries of the form: “If the future market regime were X instead of Y, how would the limit order book evolve?” We introduce the first systematic evaluation framework for counterfactual LOB generation consisting of three criteria: (1) Realism, measuring how well generated trajectories can reproduce marginal distributions, temporal dependence structure and regime variables; (2) Counterfactual validity, testing whether interventions on future regimes induce consistent changes in the generated LOB dynamics; (3) Counterfactual usefulness, assessing whether synthetic counterfactual trajectories improve downstream prediction of future market regimes.
Deep learning has achieved strong results in modeling sequential data, including event sequences, temporal point processes, and irregular time series. Recently, transformers have largely replaced recurrent networks in these tasks. However, transformers often underperform recurrent networks in classification tasks that aim to predict future targets, such as churn, user reactions, or treatment response. The reason behind this performance gap remains largely underexplored. In this paper, we identify a key limitation of transformers: the lack of a single vector representation that compactly summarizes the evolving state of a sequence. We further show that commonly used contrastive embeddings are poorly suited to capturing the local context needed for accurate forward-looking prediction. To address these challenges, we introduce history tokens, a novel concept that enables the accumulation of historical information during next-token prediction pretraining. Our approach significantly improves transformer-based models, achieving impressive results in finance, e-commerce, and healthcare tasks. The code is publicly available: https://github.com/ivan-chai/pretpp.
Reusing and migrating analog circuit intellectual property (IP) across process nodes poses a significant challenge in modern chip design. Efficient and generalizable circuit performance prediction methods for analog circuits are crucial to achieving this goal. Current data-driven approaches typically rely on manually designed features, which perform poorly on unseen circuit architectures and struggle to model the inherent structural relationships within analog designs. To address these challenges, we propose ACCFormer, a novel topology-aware Transformer framework for predicting performance metrics of analog circuit. Our model combines device parameters with connectivity data to learn topology-aware representations, followed by a performance-oriented cross-attention mechanism where trainable metric queries adaptively focus on the most critical devices for each target parameter. Validated across different process nodes, our model achieves state-of-the-art prediction accuracy and demonstrates strong cross-process adaptability, highlighting its potential to accelerate IP reuse and reduce design cycles.
Brain connectivity analysis is a fundamental tool for identifying biomarkers and understanding of neurological disorders. Most existing approaches employ graph transformers over undirected functional connectivity networks, which are typically estimated using correlation statistics. Although effective for capturing statistical associations, these models do not represent directed interactions between brain regions that arise from causal relationships. As a result, direction-specific disease mechanisms are not explicitly modeled, and interpretability is often limited. To address this gap, we present BrainCGT, a brain graph transformer designed to model causal connectivity inferred from fMRI time-series data. In this framework, brain networks are modeled as directed graphs with a modular organization, where nodes correspond to individual brain regions and directed edges reflect causal flow of information between them. Direction-aware node representations together with direction-biased attention mechanisms allow the model to capture asymmetric interactions across regions. Experimental results on three large-scale fMRI datasets demonstrate that BrainCGT achieves consistently better performance than existing graph-based methods for neurological disorder classification. In addition, examination of the learned attention structures shows correspondence with established neurobiological pathways, suggesting improved interpretability. These results highlight the importance of incorporating causal directionality into brain graph transformer architectures for robust and interpretable neuroimaging analysis.
Chlor-alkali production is a large-scale industrial process whose operating conditions and equipment states evolve over time. Its process optimization requires ongoing trade-offs among conflicting objectives such as product yield, energy consumption, and equipment life. Existing optimization approaches are typically static and must be re-optimized after environmental changes, limiting their real-world applicability. In this work, we model the problem as a dynamic multi-objective sequential decision-making problem that continuously tracks a time-varying Pareto set under changing conditions. We propose MORL-CA, a multi-objective reinforcement learning framework that integrates offline pretraining on historical data with constrained online policy refinement. MORL-CA introduces a state-aware adaptive objective weighting mechanism within a multi-critic actor-critic architecture, enabling localized Pareto-improving policy updates while satisfying operational and safety constraints. Extensive experiments in an environment conducted from real chlor-alkali data demonstrate that MORL-CA achieves superior Pareto solution quality and smoother adaptation to dynamics compared with state-of-the-art multi-objective optimizers and MORL baselines.
Macro placement is a pivotal stage in VLSI physical design, fundamentally determining the overall chip performance. Recent data-driven placement methods have demonstrated significant potential, yet they often struggle to handle sequential dependencies and to balance topological connectivity with physical constraints. To bridge this gap, we propose MacroDiff+, a physics-guided geometric diffusion framework. Specifically, we design a dual-domain denoising architecture that couples topological connectivity encoded by heterogeneous GNNs with global geometric context modeled by a Transformer. Furthermore, we introduce Physics-Guided Sampling, an inference strategy that actively steers the generation using explicit gradients to ensure both statistical plausibility and physical validity. On the ISPD2005 MMS benchmarks, MacroDiff+ outperforms state-of-the-art baselines with a 6.1–6.2% reduction in wirelength. Notably, it exhibits superior stability and scalability on large-scale designs where prior methods fail to converge. The source code is provided at https://github.com/jhy00n/MacroDiff-plus.
Fully autonomous IP-to-Root penetration testing remains challenging for LLM agents. We conduct an exploratory study on 10 LLMs and introduce AutoPentest-Bench, an end-to-end benchmark with 13 VulnHub targets and 93 sub-tasks. From 130 interaction logs, we identify three challenges: Rigid Strategy, Contextual Forgetting, and Command Generation Hallucination. To address them, we propose PentestLLMAgent, which integrates a Task Dependency Graph (TDG) for dynamic planning and backtracking; a Hierarchical Multi-Agent Architecture (HMA) with function-calling-based tool invocation, output filtering, and semantic compression, and Executable Knowledge-Guided Command Generation (EKG-CG) for retrieving and executing pre-validated, environment-compatible commands. Evaluations demonstrate strong effectiveness: on end-to-end AutoPentest-Bench, PentestLLMAgent achieves a 77% success rate; on AutoPT’s web exploitation benchmark, it attains a 95% overall pass rate; and on the privilege escalation benchmark, it achieves a 100% success rate. In realistic end-to-end runs, it averages 10.9 minutes, 36 interaction rounds, and 69.6K tokens per target. The code, benchmark, and executable knowledge base are publicly available at https://github.com/sanbai123/PentestLLMAgent_code-and-videos.
Active inverse source localization and characterization (ISLC) in dynamic fields requires sequential decision making under partial observability, where a mobile sensor must infer latent source parameters from sparse, noisy readings. We introduce a belief-contraction-driven approach that unifies inference, stopping, and control. An attention-augmented particle filter stabilizes Bayesian belief updates through ESS-based resampling, feature-aware sparse attention smoothing, and Metropolis–Hastings rejuvenation that preserves the filtering posterior. Belief contraction (posterior dispersion) defines both a termination rule and a goal-aligned intrinsic reward, enabling reinforcement learning without distance-to-source shaping. Across seven field modalities, spatial out-of-distribution tests, and nonstationary source shifts, our agent (ATT-PFRL) achieves higher completion, faster convergence, and more accurate localization than planning and RL+Bayes baselines under similar computation. Fixed-trajectory studies also show improved ESS and lower RMSE, isolating the benefit of the inference layer.
Agricultural landscape segmentation in the Global South is challenging as it is characterized by fragmented plots, high intra-class variance, and a scarcity of labeled training data. Recent advances in segmentation have been made by Multimodal Large Language Models (MLLMs). However, current approaches encounter critical context length bottlenecks and a domain alignment gap in understanding satellite features. We address these limitations through MAgSeg, a novel, decoder-free MLLM segmentation approach. MAgSeg is an architecturally efficient approach that enables standard MLLMs to perform segmentation of complex smallholder agricultural landscapes from high-resolution satellite imagery, without requiring auxiliary vision decoders. We introduce a novel instruction tuning data format designed to enable scalable fine-tuning and post-training on high resolution satellite imagery, which enables MAgSeg to learn from the global context of the image while generating text tokens for only a patch within the image. Extensive evaluations on datasets spanning three countries in the Global South demonstrate that MAgSeg significantly outperforms state-of-the-art MLLM baselines, offering a scalable solution to map smallholder agricultural environments.
The advancement of data privacy regulations has spurred the development of Machine Unlearning (MU), which is designed to remove the influence of sensitive data from a trained model and results in an unlearned model (ULM). Despite rapid progress in MU techniques, their vulnerabilities remain underexplored, which poses risks due to potential leakage of unlearned information. In realistic scenarios, ULMs always need to be incrementally trained with the newly collected data samples, which can lead to the consequences of recalling sensitive information if the new dataset contains similar or even the same unlearned samples. To address this issue, we devise a Durable Unlearning Enhancement (DUE) framework to avoid restoring unwanted sensitive information from incremental training data samples. The DUE framework has three key components that identify sensitive samples and suppress their gradients to update ULMs. Extensive experiments on state-of-the-art MU methods across multiple real-world datasets show that the proposed DUE framework can effectively nullify the recall of sensitive information after MU, and even improve the performance of ULMs. Consequently, our work establishes a new fundamental research direction in safe training against MU vulnerabilities.
Cutting planes are central to mixed-integer linear programming (MILP) solving, yet their effectiveness hinges on expert tuning of separator configurations and hard-crafted cut-selection heuristics, creating a high barrier for non-specialists. Learning-based methods can reduce manual effort, but typically require large training datasets and often generalize poorly beyond the instance classes they are trained on. We propose an LLM-guided cutting-plane management framework that integrates large language models into the MILP solving pipeline. First, using chain-of-thought (CoT) prompting, the LLM infers an instance-specific separator configuration, deciding which separators to activate and how to set key parameters from the problem type and structural features. Then, it translates the evolving branch-and-bound state in natural language to perform stage-aware cut selection, choosing a high-quality subset of cuts that tightens the relaxation and improves overall solver performance. Leveraging the LLM's reasoning capabilities and rich background knowledge, our work removes dependence on domain-specific training data and substantially reduces reliance on expert-crafted configurations. Experiments show consistent improvements over SCIP's default settings, hard-crafted heuristics, and recent learning-based cut selection baselines.
Traffic forecasting task is significant to modern urban management. Recently, there is growing attention on large-scale forecasting, as it better reflects the complexity of real-world traffic networks. However, existing models often exhibit quadratic computational complexity, making them impractical for large-scale real-world scenarios. In this paper, we propose a novel framework, Spatio-Temporal Hierarchical Mixer (HieraMix), which leverages an all-MLP architecture for efficient and effective large-scale traffic forecasting. HieraMix employs a hierarchical spatiotemporal mixing block to extract multi-resolution features through bottom-up aggregation and top-down propagation. Furthermore, an adaptive region mixer generates transformation matrices based on regional semantics, enabling our model to dynamically capture evolving spatiotemporal patterns for different regions. Extensive experiments conducted on four large-scale real-world datasets demonstrate that the proposed method not only achieves state-of-the-art performance but also exhibits competitive computational efficiency.
Detecting spoofing in financial trading is a critical data mining task. While traditional machine learning models focus on individual node features, graph-based methodologies have shown superior performance by integrating relational data and structure information. However, spoofing transactions often exhibit distribution shifts, making historical data less effective. Instead, certain consistent trading patterns, such as motif structures, remain robust for analysis across different distributions. Therefore, this paper introduces the Structure-Augmented Generative Graph Model (SAG^2M) to detect conspiracy spoofing through substructure frequency-augmented detection. Our approach extracts subgraph pattern frequencies among neighboring nodes using an enumeration algorithm, and encodes these motif frequencies into a structure-augmented generative framework. A temporal and heterogeneous graph generation and aggregation scheme is then applied to collect neighborhood node information, effectively uncovering conspiracy spoofing patterns. Experiments on benchmark datasets demonstrate that SAG^2M outperforms existing models in detection accuracy. Case studies further highlight its superior effectiveness in identifying complex fraudulent behaviors.
Due to the high value and high failure rates of startups, predicting their success is a critical challenge. Existing approaches typically model startup success from a single decision-maker's perspective, overlooking the collective dynamics that dominate real-world venture capital (VC) decision-making. We propose SimVC-CAS, a collective agent system that simulates VC decisions as a multi-agent interaction process. By designing role-playing agents and a GNN-based supervised interaction module, we reformulate startup financing prediction as a group decision-making task, capturing both enterprise fundamentals and investor network dynamics. Each agent represents an investor with distinct traits and preferences, enabling heterogeneous evaluations and realistic information exchange over a graph-structured co-investment network. Using both proprietary and public VC data with strict anti-leakage controls, we show that SimVC-CAS significantly improves predictive performance, achieving approximately 25% relative improvement in average precision@10, while exhibiting consistency with real investor decisions. The interaction mechanism is particularly effective for network-central startups, confirming the importance of network in VC decision-making. Analysis of agents' reasoning for decision changes further reveals how network environment influence decision quality, demonstrating the system's interpretability. Our approach may generalize to broader group decision-making scenarios. Our code is available at https://github.com/ZhangDataLab/SimVC-CAS.