From Correctness to Utility: Gain-Based Prefix Evaluation for LLM Reasoning

Reasoning prefixes shape the future trajectory of LLM problem solving, yet existing process reward models usually evaluate them through local step correctness. We argue that correctness is a useful but indirect proxy for the effect we ultimately care about: whether a prefix increases the probability of successful completion. We define this effect as prefix gain, the solve-rate improvement induced by conditioning lightweight student model group on a prefix, and use it to train a Prefix Utility Model (PUM) with a simple pairwise ranking objective. PUM learns outcome-grounded prefix utility and can score both complete trajectories and partial reasoning prefixes. Across Best-of-$N$ selection, beam search, and reinforcement learning on mathematical reasoning, PUM provides a strong prefix-level supervision signal, especially when candidate pools are large, search budgets increase, or rule-based rewards are sparse. We release all data, models, and code at https://zhiqix.github.io/pum-project-page.

Mind the Tool Failures: Achieving Synergistic Tool Gains for Medical Agents

Medical AI agents increasingly use external tools for diagnosis, treatment recommendation, and evidence retrieval, yet most existing approaches assume that task-appropriate tools are reliable within their intended scope. This assumption is fragile in real clinical settings, where even relevant tools may fail on challenging instances and lead to unsafe downstream decisions. To address this issue, we study medical tool use under imperfect-tool settings to correct failure instances missed by individual tools. Instance-dependent failure patterns create a gap between the best fixed single tool and an ideal instance-wise selector, which we refer to as the Single-Oracle risk gap. The core challenge is that conventional task-level tool selection cannot realize this gap, as it is inherently bounded by the performance of the best single tool. Motivated by this observation, we therefore account for instance-level heterogeneity and formulate tool use as an instance-level selection problem. Particularly, we propose a GRPO-based reinforcement learning framework with rewards for probabilistic risk minimization and disagreement-aware synergy learning, which promotes instance-level correction of erroneous tool consensus. Furthermore, an entropy-guided sampling strategy is adopted to upweight high-disagreement instances, which provide stronger signals for learning instance-specific tool synergy. These two components complement each other in mitigating instance-level heterogeneity and improving tool synergy. Experiments on two tasks and seven medical benchmarks show that our method consistently achieves robust and stable improvements over a broad range of baselines, highlighting the importance of synergy-aware tool use for reliable medical agentic systems.

TransXion: A High-Fidelity Graph Benchmark for Realistic Anti-Money Laundering

Money laundering poses severe risks to global financial systems, driving the widespread adoption of machine learning for transaction monitoring. However, progress remains stifled by the lack of realistic benchmarks. Existing transaction-graph datasets suffer from two pervasive limitations: (i) they provide sparse node-level semantics beyond anonymized identifiers, and (ii) they rely on template-driven anomaly injection, which biases benchmarks toward static structural motifs and yields overly optimistic assessments of model robustness. We propose TransXion, a benchmark ecosystem for Anti-Money Laundering (AML) research that integrates profile-aware simulation of normal activity with stochastic, non-template synthesis of illicit subgraphs.TransXion jointly models persistent entity profiles and conditional transaction behavior, enabling evaluation of "out-of-character" anomalies where observed activity contradicts an entity's socio-economic context. The resulting dataset comprises approximately 3 million transactions among 50,000 entities, each endowed with rich demographic and behavioral attributes. Empirical analyses show that TransXion reproduces key structural properties of payment networks, including heavy-tailed activity distributions and localized subgraph structure. Across a diverse array of detection models spanning multiple algorithmic paradigms, TransXion yields substantially lower detection performance than widely used benchmarks, demonstrating increased difficulty and realism. TransXion provides a more faithful testbed for developing context-aware and robust AML detection methods. The dataset and code are publicly available at https://github.com/chaos-max/TransXion.

GAM-RAG: Gain-Adaptive Memory for Evolving Retrieval in Retrieval-Augmented Generation

Retrieval-Augmented Generation (RAG) grounds large language models with external evidence, but many implementations rely on pre-built indices that remain static after construction. Related queries therefore repeat similar multi-hop traversal, increasing latency and compute. Motivated by schema-based learning in cognitive neuroscience, we propose GAM-RAG, a training-free framework that accumulates retrieval experience from recurring or related queries and updates retrieval memory over time. GAM-RAG builds a lightweight, relation-free hierarchical index whose links capture potential co-occurrence rather than fixed semantic relations. During inference, successful retrieval episodes provide sentence-level feedback, updating sentence memories so evidence useful for similar reasoning types becomes easier to activate later. To balance stability and adaptability under noisy feedback, we introduce an uncertainty-aware, Kalman-inspired gain rule that jointly updates memory states and perplexity-based uncertainty estimates. It applies fast updates for reliable novel signals and conservative refinement for stable or noisy memories. We provide a theoretical analysis of the update dynamics, and empirically show that GAM-RAG improves average performance by 3.95% over the strongest baseline and by 8.19% with 5-turn memory, while reducing inference cost by 61%. Our code and datasets are available at: https://anonymous.4open.science/r/GAM_RAG-2EF6.

Behavioral Consistency Validation for LLM Agents: An Analysis of Trading-Style Switching through Stock-Market Simulation

Recent works have increasingly applied Large Language Models (LLMs) as agents in financial stock market simulations to test if micro-level behaviors aggregate into macro-level phenomena. However, a crucial question arises: Do LLM agents' behaviors align with real market participants? This alignment is key to the validity of simulation results. To explore this, we select a financial stock market scenario to test behavioral consistency. Investors are typically classified as fundamental or technical traders, but most simulations fix strategies at initialization, failing to reflect real-world trading dynamics. In this work, we assess whether agents' strategy switching aligns with financial theory, providing a framework for this evaluation. We operationalize four behavioral-finance drivers-loss aversion, herding, wealth differentiation, and price misalignment-as personality traits set via prompting and stored long-term. In year-long simulations, agents process daily price-volume data, trade under a designated style, and reassess their strategy every 10 trading days. We introduce four alignment metrics and use Mann-Whitney U tests to compare agents' style-switching behavior with financial theory. Our results show that recent LLMs' switching behavior is only partially consistent with behavioral-finance theories, highlighting the need for further refinement in aligning agent behavior with financial theory.

Rethinking the Role of Entropy in Optimizing Tool-Use Behaviors for Large Language Model Agents

Tool-using agents based on Large Language Models (LLMs) excel in tasks such as mathematical reasoning and multi-hop question answering. However, in long trajectories, agents often trigger excessive and low-quality tool calls, increasing latency and degrading inference performance, making managing tool-use behavior challenging. In this work, we conduct entropy-based pilot experiments and observe a strong positive correlation between entropy reduction and high-quality tool calls. Building on this finding, we propose using entropy reduction as a supervisory signal and design two reward strategies to address the differing needs of optimizing tool-use behavior. Sparse outcome rewards provide coarse, trajectory-level guidance to improve efficiency, while dense process rewards offer fine-grained supervision to enhance performance. Experiments across diverse domains show that both reward designs improve tool-use behavior: the former reduces tool calls by 72.07% compared to the average of baselines, while the latter improves performance by 22.27%. These results position entropy reduction as a key mechanism for enhancing tool-use behavior, enabling agents to be more adaptive in real-world applications.

Schrödinger's Navigator: Imagining an Ensemble of Futures for Zero-Shot Object Navigation

Zero-shot object navigation (ZSON) requires a robot to locate a target object in a previously unseen environment without relying on pre-built maps or task-specific training. However, existing ZSON methods often struggle in realistic and cluttered environments, particularly when the scene contains heavy occlusions, unknown risks, or dynamically moving target objects. To address these challenges, we propose \textbf{Schrödinger's Navigator}, a navigation framework inspired by Schrödinger's thought experiment on uncertainty. The framework treats unobserved space as a set of plausible future worlds and reasons over them before acting. Conditioned on egocentric visual inputs and three candidate trajectories, a trajectory-conditioned 3D world model imagines future observations along each path. This enables the agent to see beyond occlusions and anticipate risks in unseen regions without requiring extra detours or dense global mapping. The imagined 3D observations are fused into the navigation map and used to update a value map. These updates guide the policy toward trajectories that avoid occlusions, reduce exposure to uncertain space, and better track moving targets. Experiments on a Go2 quadruped robot across three challenging scenarios, including severe static occlusions, unknown risks, and dynamically moving targets, show that Schrödinger's Navigator consistently outperforms strong ZSON baselines in self-localization, object localization, and overall Success Rate in occlusion-heavy environments. These results demonstrate the effectiveness of trajectory-conditioned 3D imagination in enabling robust zero-shot object navigation.

How AI Agents Follow the Herd of AI? Network Effects, History, and Machine Optimism

Understanding decision-making in multi-AI-agent frameworks is crucial for analyzing strategic interactions in network-effect-driven contexts. This study investigates how AI agents navigate network-effect games, where individual payoffs depend on peer participatio--a context underexplored in multi-agent systems despite its real-world prevalence. We introduce a novel workflow design using large language model (LLM)-based agents in repeated decision-making scenarios, systematically manipulating price trajectories (fixed, ascending, descending, random) and network-effect strength. Our key findings include: First, without historical data, agents fail to infer equilibrium. Second, ordered historical sequences (e.g., escalating prices) enable partial convergence under weak network effects but strong effects trigger persistent "AI optimism"--agents overestimate participation despite contradictory evidence. Third, randomized history disrupts convergence entirely, demonstrating that temporal coherence in data shapes LLMs' reasoning, unlike humans. These results highlight a paradigm shift: in AI-mediated systems, equilibrium outcomes depend not just on incentives, but on how history is curated, which is impossible for human.

Reconstruction as a Bridge for Event-Based Visual Question Answering

Integrating event cameras with Multimodal Large Language Models (MLLMs) promises general scene understanding in challenging visual conditions, yet requires navigating a trade-off between preserving the unique advantages of event data and ensuring compatibility with frame-based models. We address this challenge by using reconstruction as a bridge, proposing a straightforward Frame-based Reconstruction and Tokenization (FRT) method and designing an efficient Adaptive Reconstruction and Tokenization (ART) method that leverages event sparsity. For robust evaluation, we introduce EvQA, the first objective, real-world benchmark for event-based MLLMs, comprising 1,000 event-Q&A pairs from 22 public datasets. Our experiments demonstrate that our methods achieve state-of-the-art performance on EvQA, highlighting the significant potential of MLLMs in event-based vision.

Teach2Eval: An Indirect Evaluation Method for LLM by Judging How It Teaches

Recent progress in large language models (LLMs) has outpaced the development of effective evaluation methods. Traditional benchmarks rely on task-specific metrics and static datasets, which often suffer from fairness issues, limited scalability, and contamination risks. In this paper, we introduce Teach2Eval, an indirect evaluation framework inspired by the Feynman Technique. Instead of directly testing LLMs on predefined tasks, our method evaluates a model's multiple abilities to teach weaker student models to perform tasks effectively. By converting open-ended tasks into standardized multiple-choice questions (MCQs) through teacher-generated feedback, Teach2Eval enables scalable, automated, and multi-dimensional assessment. Our approach not only avoids data leakage and memorization but also captures a broad range of cognitive abilities that are orthogonal to current benchmarks. Experimental results across 26 leading LLMs show strong alignment with existing human and model-based dynamic rankings, while offering additional interpretability for training guidance.