PolicySim: An LLM-Based Agent Social Simulation Sandbox for Proactive Policy Optimization

Social platforms serve as central hubs for information exchange, where user behaviors and platform interventions jointly shape opinions. However, intervention policies like recommendation and content filtering, can unintentionally amplify echo chambers and polarization, posing significant societal risks. Proactively evaluating the impact of such policies is therefore crucial. Existing approaches primarily rely on reactive online A/B testing, where risks are identified only after deployment, making risk identification delayed and costly. LLM-based social simulations offer a promising pre-deployment alternative, but current methods fall short in realistically modeling platform interventions and incorporating feedback from the platform. Bridging these gaps is essential for building actionable frameworks to assess and optimize platform policies. To this end, we propose PolicySim, an LLM-based social simulation sandbox for the proactive assessment and optimization of intervention policies. PolicySim models the bidirectional dynamics between user behavior and platform interventions through two key components: (1) a user agent module refined via supervised fine-tuning (SFT) and direct preference optimization (DPO) to achieve platform-specific behavioral realism; and (2) an adaptive intervention module that employs a contextual bandit with message passing to capture dynamic network structures. Experiments show that PolicySim can accurately simulate platform ecosystems at both micro and macro levels and support effective intervention policy.

RAPO: Expanding Exploration for LLM Agents via Retrieval-Augmented Policy Optimization

Agentic Reinforcement Learning (Agentic RL) has shown remarkable potential in large language model-based (LLM) agents. These works can empower LLM agents to tackle complex tasks via multi-step, tool-integrated reasoning. However, an inherent limitation of existing Agentic RL methods is their reliance on a pure on-policy paradigm for exploration, restricting exploration to the agent's self-generated outputs and preventing the discovery of new reasoning perspectives for further improvement. While recent efforts incorporate auxiliary off-policy signals to enhance exploration, they typically utilize full off-policy trajectories for trajectory-level policy estimation, overlooking the necessity for the fine-grained, step-level exploratory dynamics within agentic rollout. In this paper, we revisit exploration in Agentic RL and propose Retrieval-Augmented Policy Optimization (RAPO), a novel RL framework that introduces retrieval to explicitly expand exploration during training. To achieve this, we decompose the Agentic RL training process into two phases: (i) Hybrid-policy Agentic Rollout, and (ii) Retrieval-aware Policy Optimization. Specifically, we propose a Hybrid-policy Agentic Rollout strategy, which allows the agents to continuously reason over the retrieved off-policy step-level traces. It dynamically extends the reasoning receptive field of agents, enabling broader exploration conditioned on external behaviors. Subsequently, we introduce the Retrieval-aware Policy Optimization mechanism, which calibrates the policy gradient estimation with retrieval reward and importance shaping, stabilizing training and prioritizing retrieval-illuminating exploration. Extensive experiments show that RAPO achieves an +5.0% average gain on fourteen datasets across three agentic reasoning tasks, while delivering 1.2x faster training efficiency.

Tree of Preferences for Diversified Recommendation

Diversified recommendation has attracted increasing attention from both researchers and practitioners, which can effectively address the homogeneity of recommended items. Existing approaches predominantly aim to infer the diversity of user preferences from observed user feedback. Nonetheless, due to inherent data biases, the observed data may not fully reflect user interests, where underexplored preferences can be overwhelmed or remain unmanifested. Failing to capture these preferences can lead to suboptimal diversity in recommendations. To fill this gap, this work aims to study diversified recommendation from a data-bias perspective. Inspired by the outstanding performance of large language models (LLMs) in zero-shot inference leveraging world knowledge, we propose a novel approach that utilizes LLMs' expertise to uncover underexplored user preferences from observed behavior, ultimately providing diverse and relevant recommendations. To achieve this, we first introduce Tree of Preferences (ToP), an innovative structure constructed to model user preferences from coarse to fine. ToP enables LLMs to systematically reason over the user's rationale behind their behavior, thereby uncovering their underexplored preferences. To guide diversified recommendations using uncovered preferences, we adopt a data-centric approach, identifying candidate items that match user preferences and generating synthetic interactions that reflect underexplored preferences. These interactions are integrated to train a general recommender for diversification. Moreover, we scale up overall efficiency by dynamically selecting influential users during optimization. Extensive evaluations of both diversity and relevance show that our approach outperforms existing methods in most cases and achieves near-optimal performance in others, with reasonable inference latency.

Can Graph Neural Networks Expose Training Data Properties? An Efficient Risk Assessment Approach

Graph neural networks (GNNs) have attracted considerable attention due to their diverse applications. However, the scarcity and quality limitations of graph data present challenges to their training process in practical settings. To facilitate the development of effective GNNs, companies and researchers often seek external collaboration. Yet, directly sharing data raises privacy concerns, motivating data owners to train GNNs on their private graphs and share the trained models. Unfortunately, these models may still inadvertently disclose sensitive properties of their training graphs (e.g., average default rate in a transaction network), leading to severe consequences for data owners. In this work, we study graph property inference attack to identify the risk of sensitive property information leakage from shared models. Existing approaches typically train numerous shadow models for developing such attack, which is computationally intensive and impractical. To address this issue, we propose an efficient graph property inference attack by leveraging model approximation techniques. Our method only requires training a small set of models on graphs, while generating a sufficient number of approximated shadow models for attacks. To enhance diversity while reducing errors in the approximated models, we apply edit distance to quantify the diversity within a group of approximated models and introduce a theoretically guaranteed criterion to evaluate each model's error. Subsequently, we propose a novel selection mechanism to ensure that the retained approximated models achieve high diversity and low error. Extensive experiments across six real-world scenarios demonstrate our method's substantial improvement, with average increases of 2.7% in attack accuracy and 4.1% in ROC-AUC, while being 6.5$\times$ faster compared to the best baseline.

Can Modifying Data Address Graph Domain Adaptation?

Graph neural networks (GNNs) have demonstrated remarkable success in numerous graph analytical tasks. Yet, their effectiveness is often compromised in real-world scenarios due to distribution shifts, limiting their capacity for knowledge transfer across changing environments or domains. Recently, Unsupervised Graph Domain Adaptation (UGDA) has been introduced to resolve this issue. UGDA aims to facilitate knowledge transfer from a labeled source graph to an unlabeled target graph. Current UGDA efforts primarily focus on model-centric methods, such as employing domain invariant learning strategies and designing model architectures. However, our critical examination reveals the limitations inherent to these model-centric methods, while a data-centric method allowed to modify the source graph provably demonstrates considerable potential. This insight motivates us to explore UGDA from a data-centric perspective. By revisiting the theoretical generalization bound for UGDA, we identify two data-centric principles for UGDA: alignment principle and rescaling principle. Guided by these principles, we propose GraphAlign, a novel UGDA method that generates a small yet transferable graph. By exclusively training a GNN on this new graph with classic Empirical Risk Minimization (ERM), GraphAlign attains exceptional performance on the target graph. Extensive experiments under various transfer scenarios demonstrate the GraphAlign outperforms the best baselines by an average of 2.16%, training on the generated graph as small as 0.25~1% of the original training graph.

Better with Less: A Data-Active Perspective on Pre-Training Graph Neural Networks

Pre-training on graph neural networks (GNNs) aims to learn transferable knowledge for downstream tasks with unlabeled data, and it has recently become an active research area. The success of graph pre-training models is often attributed to the massive amount of input data. In this paper, however, we identify the curse of big data phenomenon in graph pre-training: more training data do not necessarily lead to better downstream performance. Motivated by this observation, we propose a better-with-less framework for graph pre-training: fewer, but carefully chosen data are fed into a GNN model to enhance pre-training. The proposed pre-training pipeline is called the data-active graph pre-training (APT) framework, and is composed of a graph selector and a pre-training model. The graph selector chooses the most representative and instructive data points based on the inherent properties of graphs as well as predictive uncertainty. The proposed predictive uncertainty, as feedback from the pre-training model, measures the confidence level of the model in the data. When fed with the chosen data, on the other hand, the pre-training model grasps an initial understanding of the new, unseen data, and at the same time attempts to remember the knowledge learned from previous data. Therefore, the integration and interaction between these two components form a unified framework (APT), in which graph pre-training is performed in a progressive and iterative way. Experiment results show that the proposed APT is able to obtain an efficient pre-training model with fewer training data and better downstream performance.