Adaptation is the cornerstone of effective collaboration among heterogeneous team members. In human-agent teams, artificial agents need to adapt to their human partners in real time, as individuals often have unique preferences and policies that may change dynamically throughout interactions. This becomes particularly challenging in tasks with time pressure and complex strategic spaces, where identifying partner behaviors and selecting suitable responses is difficult. In this work, we introduce a strategy-conditioned cooperator framework that learns to represent, categorize, and adapt to a broad range of potential partner strategies in real-time. Our approach encodes strategies with a variational autoencoder to learn a latent strategy space from agent trajectory data, identifies distinct strategy types through clustering, and trains a cooperator agent conditioned on these clusters by generating partners of each strategy type. For online adaptation to novel partners, we leverage a fixed-share regret minimization algorithm that dynamically infers and adjusts the partner's strategy estimation during interaction. We evaluate our method in a modified version of the Overcooked domain, a complex collaborative cooking environment that requires effective coordination among two players with a diverse potential strategy space. Through these experiments and an online user study, we demonstrate that our proposed agent achieves state of the art performance compared to existing baselines when paired with novel human, and agent teammates.

Automated red teaming frameworks for Large Language Models (LLMs) have become increasingly sophisticated, yet they share a fundamental limitation: their jailbreak logic is confined to selecting, combining, or refining pre-existing attack strategies. This binds their creativity and leaves them unable to autonomously invent entirely new attack mechanisms. To overcome this gap, we introduce \textbf{EvoSynth}, an autonomous framework that shifts the paradigm from attack planning to the evolutionary synthesis of jailbreak methods. Instead of refining prompts, EvoSynth employs a multi-agent system to autonomously engineer, evolve, and execute novel, code-based attack algorithms. Crucially, it features a code-level self-correction loop, allowing it to iteratively rewrite its own attack logic in response to failure. Through extensive experiments, we demonstrate that EvoSynth not only establishes a new state-of-the-art by achieving an 85.5\% Attack Success Rate (ASR) against highly robust models like Claude-Sonnet-4.5, but also generates attacks that are significantly more diverse than those from existing methods. We release our framework to facilitate future research in this new direction of evolutionary synthesis of jailbreak methods. Code is available at: https://github.com/dongdongunique/EvoSynth.

Deploying embodied agents that can answer questions about their surroundings in realistic real-world settings remains difficult, partly due to the scarcity of benchmarks that faithfully capture practical operating conditions. We propose infrastructure inspection as a compelling domain for open-vocabulary Embodied Question Answering (EQA): it naturally demands multi-scale reasoning, long-range spatial understanding, and complex semantic relationships, while offering unique evaluation advantages via standardized National Bridge Inventory (NBI) condition ratings (0-9), professional inspection reports, and egocentric imagery. We introduce BridgeEQA, a benchmark of 2,200 open-vocabulary question-answer pairs (in the style of OpenEQA) grounded in professional inspection reports across 200 real-world bridge scenes with 47.93 images on average per scene. Questions require synthesizing visual evidence across multiple images and aligning responses with NBI condition ratings. We further propose a new EQA metric Image Citation Relevance to evaluate the ability of a model to cite relevant images. Evaluations of state-of-the-art vision-language models reveal substantial performance gaps under episodic memory EQA settings. To address this, we propose Embodied Memory Visual Reasoning (EMVR), which formulates inspection as sequential navigation over an image-based scene graph: images are nodes, and an agent takes actions to traverse views, compare evidence, and reason within a Markov decision process. EMVR shows strong performance over the baselines. We publicly release both the dataset and code.

Notice to Air Missions (NOTAMs) serve as a critical channel for disseminating key flight safety information, yet their complex linguistic structures and implicit reasoning pose significant challenges for automated parsing. Existing research mainly focuses on surface-level tasks such as classification and named entity recognition, lacking deep semantic understanding. To address this gap, we propose NOTAM semantic parsing, a task emphasizing semantic inference and the integration of aviation domain knowledge to produce structured, inference-rich outputs. To support this task, we construct Knots (Knowledge and NOTAM Semantics), a high-quality dataset of 12,347 expert-annotated NOTAMs covering 194 Flight Information Regions, enhanced through a multi-agent collaborative framework for comprehensive field discovery. We systematically evaluate a wide range of prompt-engineering strategies and model-adaptation techniques, achieving substantial improvements in aviation text understanding and processing. Our experimental results demonstrate the effectiveness of the proposed approach and offer valuable insights for automated NOTAM analysis systems. Our code is available at: https://github.com/Estrellajer/Knots.

ABSTRACT Large language models (LLMs) have shown strong reasoning capabilities and are increasingly explored for financial trading. Existing LLM-based trading agents, however, largely focus on single-step prediction and lack integrated mechanisms for risk management, which reduces their effectiveness in volatile markets. We introduce FinRS, a risk-sensitive trading framework that combines hierarchical market analysis, dual-decision agents, and multi-timescale reward reflection to align trading actions with both return objectives and downside risk constraints. Experiments on multiple stocks and market conditions show that FinRS achieves superior profitability and stability compared to state-of-the-art methods. Index T erms-- Trading Agent, Risk-Sensitive, Real Markets 1. INTRODUCTION In recent years, large language models (LLMs) [1, 2] have demonstrated significant potential in financial trading.

Task-oriented dialogue systems aim to accomplish various user goals through natural language communication, which often involve complexity and require multiple dialogue turns to complete [1-3]. For instance, when assisting users in booking air tickets, a task-oriented dialogue system engages in a conversation to gather information such as the departure place, destination, and departure time. Once sufficient information is obtained, the system automatically handles the booking process. The convenience offered by this natural language interaction has led to a growing interest in task-oriented dialogue systems in recent years [4-6]. Traditionally, task-oriented dialogue systems are generally modeled as intelligent agents that have access to back-end APIs to acquire knowledge in a database [7-9], thereby using this knowledge to help users complete various tasks. These agents follow a pipeline process in the dialogue with users: predict the user's intention, extract slot values in the user's utterance, call API to access the database and response to the user [10-15]. For example, as shown in Figure 1, when a user desires to book a restaurant, the agent engages in a dialogue where, in the first 4 turns, the user seeks a restaurant meeting specific requirements, prompting the agent to call the "find_restaurant" API. In the last 2 turns, the user provides detailed reservation information, leading to the agent calling the "book_restaurant" API. However, in real-world scenarios, the availability of customized APIs for building practical task-oriented dialogue systems is limited, primarily due to two reasons.

Autonomous agents operating in sequential decision-making tasks under uncertainty can benefit from external action suggestions, which provide valuable guidance but inherently vary in reliability. Existing methods for incorporating such advice typically assume static and known suggester quality parameters, limiting practical deployment. We introduce a framework that dynamically learns and adapts to varying suggester reliability in partially observable environments. First, we integrate suggester quality directly into the agent's belief representation, enabling agents to infer and adjust their reliance on suggestions through Bayesian inference over suggester types. Second, we introduce an explicit ``ask'' action allowing agents to strategically request suggestions at critical moments, balancing informational gains against acquisition costs. Experimental evaluation demonstrates robust performance across varying suggester qualities, adaptation to changing reliability, and strategic management of suggestion requests. This work provides a foundation for adaptive human-agent collaboration by addressing suggestion uncertainty in uncertain environments.

Despite the explosive growth of AI and the technologies built upon it, predicting and inferring the sub-optimal behavior of users or human collaborators remains a critical challenge. In many cases, such behaviors are not a result of irrationality, but rather a rational decision made given inherent cognitive bounds and biased beliefs about the world. In this paper, we formally introduce a class of computational-rational (CR) user models for cognitively-bounded agents acting optimally under biased beliefs. The key novelty lies in explicitly modeling how a bounded memory process leads to a dynamically inconsistent and biased belief state and, consequently, sub-optimal sequential decision-making. We address the challenge of identifying the latent user-specific bound and inferring biased belief states from passive observations on the fly. We argue that for our formalized CR model family with an explicit and parameterized cognitive process, this challenge is tractable. To support our claim, we propose an efficient online inference method based on nested particle filtering that simultaneously tracks the user's latent belief state and estimates the unknown cognitive bound from a stream of observed actions. We validate our approach in a representative navigation task using memory decay as an example of a cognitive bound. With simulations, we show that (1) our CR model generates intuitively plausible behaviors corresponding to different levels of memory capacity, and (2) our inference method accurately and efficiently recovers the ground-truth cognitive bounds from limited observations ($\le 100$ steps). We further demonstrate how this approach provides a principled foundation for developing adaptive AI assistants, enabling adaptive assistance that accounts for the user's memory limitations.

We consider a variant of the target defense problem where a single defender is tasked to capture a sequence of incoming intruders. Both the defender and the intruders have non-holonomic dynamics. The intruders' objective is to breach the target perimeter without being captured by the defender, while the defender's goal is to capture as many intruders as possible. After one intruder breaches or is captured, the next appears randomly on a fixed circle surrounding the target. Therefore, the defender's final position in one game becomes its starting position for the next. We divide an intruder-defender engagement into two phases, partial information and full information, depending on the information available to the players. We address the capturability of an intruder by the defender using the notions of Dubins path and guarding arc. We quantify the percentage of capture for both finite and infinite sequences of incoming intruders. Finally, the theoretical results are verified through numerical examples using Monte-Carlo-type random trials of experiments.

Particle filters (PFs) are often combined with swarm intelligence (SI) algorithms, such as Chicken Swarm Optimization (CSO), for particle rejuvenation. Separately, Kullback--Leibler divergence (KLD) sampling is a common strategy for adaptively sizing the particle set. However, the theoretical interaction between SI-based rejuvenation kernels and KLD-based adaptive sampling is not yet fully understood. This paper investigates this specific interaction. We analyze, under a simplified modeling framework, the effect of the CSO rejuvenation step on the particle set distribution. We propose that the fitness-driven updates inherent in CSO can be approximated as a form of mean-square contraction. This contraction tends to produce a particle distribution that is more concentrated than that of a baseline PF, or in mathematical terms, a distribution that is plausibly more ``peaked'' in a majorization sense. By applying Karamata's inequality to the concave function that governs the expected bin occupancy in KLD-sampling, our analysis suggests a connection: under the stated assumptions, the CSO-enhanced PF (CPF) is expected to require a lower \emph{expected} particle count than the standard PF to satisfy the same statistical error bound. The goal of this study is not to provide a fully general proof, but rather to offer a tractable theoretical framework that helps to interpret the computational efficiency empirically observed when combining these techniques, and to provide a starting point for designing more efficient adaptive filters.