Large Language Model (LLM)-based agentic systems have shown strong capabilities across various tasks. However, existing multi-agent frameworks often rely on static or task-level workflows, which either over-process simple queries or underperform on complex ones, while also neglecting the efficiency-performance trade-offs across heterogeneous LLMs. To address these limitations, we propose Difficulty-Aware Agentic Orchestration (DAAO), which can dynamically generate query-specific multi-agent workflows guided by predicted query difficulty. DAAO comprises three interdependent modules: a variational autoencoder (VAE) for difficulty estimation, a modular operator allocator, and a cost- and performance-aware LLM router. A self-adjusting policy updates difficulty estimates based on workflow success, enabling simpler workflows for easy queries and more complex strategies for harder ones. Experiments on six benchmarks demonstrate that DAAO surpasses prior multi-agent systems in both accuracy and inference efficiency, validating its effectiveness for adaptive, difficulty-aware reasoning. Our code is open-sourced at https://github.com/AutoAgents-ai/DAAO

Detecting fraudulent behavior across diverse domains presents a significant challenge due to the adaptive and elusive activities of fraud agents. Furthermore, imbalanced data distributions and limited labeled examples increase the difficulty of detecting fraud agents. To address these challenges, we propose Catch Me If You Can—a Multi-Agent Framework to generate synthetic datasets and simulate various types of fraudulent behavior, including but not limited to anti-money laundering (AML), credit card fraud, bot attacks, and malicious traffic. Our framework comprises two core agent types: (1) Detectors, trained to identify suspicious patterns in scenarios, and (2) Transaction Agents, including both legitimate participants and adversarial fraud agents employing strategies to evade detection. In this framework, detectors iteratively refine their detection strategies while fraud agents evolve adaptive tactics to disguise illicit activities, creating an adversarial coevolutionary environment. This dynamic fosters the generation of high-dimensional and realistic datasets for training and testing. By integrating synthetic pre-training with transfer learning, the framework leverages a variety of real-world datasets—including IEEE-CIS Fraud Detection, Credit Card Fraud Detection, and Elliptic++—demonstrating its broad applicability across multiple fraud domains. Our approach significantly improves detection performance, bridging the gap between simulation and real-world applications. It enables robust training across heterogeneous fraud behaviors, contributing to the development of resilient, generalizable solutions for financial security and fraud prevention.

Recently, large language models (LLMs) have achieved significant progress in automated code generation. Despite their strong instruction-following capabilities, these models frequently struggled to align with user intent in the coding scenario. In particular, they were hampered by datasets that lacked diversity and failed to address specialized tasks or edge cases. Furthermore, challenges in supervised fine-tuning (SFT) and reinforcement learning from human feedback (RLHF) led to failures in generating precise, human-intent-aligned code. To tackle these challenges and improve the code generation performance for automated programming systems, we propose Feedback-driven Adaptive Long/short-term memory reinforced Coding OptimizatioN (i.e., FALCON). FALCON leverages long-term memory to retain and apply learned knowledge, short-term memory to incorporate immediate feedback, and meta-reinforcement learning with feedback rewards to address global-local bi-level optimization and enhance adaptability across diverse code generation tasks. Extensive experiments show that FALCON achieves state-of-the-art performance, outperforming other reinforcement learning methods by over 4.5% on MBPP and 6.1% on Humaneval, with the code publicly available. https://anonymous.4open.science/r/FALCON-3B64/README.md.

Modern text-to-image generation systems have enabled the creation of remarkably realistic and high-quality visuals, yet they often falter when handling the inherent ambiguities in user prompts. In this work, we present Twin-Co, a framework that leverages synchronized, co-adaptive dialogue to progressively refine image generation. Instead of a static generation process, Twin-Co employs a dynamic, iterative workflow where an intelligent dialogue agent continuously interacts with the user. Initially, a base image is generated from the user's prompt. Then, through a series of synchronized dialogue exchanges, the system adapts and optimizes the image according to evolving user feedback. The co-adaptive process allows the system to progressively narrow down ambiguities and better align with user intent. Experiments demonstrate that Twin-Co not only enhances user experience by reducing trial-and-error iterations but also improves the quality of the generated images, streamlining creative process across various applications.