Clinical decision-making in radiology increasingly benefits from artificial intelligence (AI), particularly through large language models (LLMs). However, traditional retrieval-augmented generation (RAG) systems for radiology question answering (QA) typically rely on single-step retrieval, limiting their ability to handle complex clinical reasoning tasks. Here we propose an agentic RAG framework enabling LLMs to autonomously decompose radiology questions, iteratively retrieve targeted clinical evidence from Radiopaedia, and dynamically synthesize evidence-based responses. We evaluated 24 LLMs spanning diverse architectures, parameter scales (0.5B to >670B), and training paradigms (general-purpose, reasoning-optimized, clinically fine-tuned), using 104 expert-curated radiology questions from previously established RSNA-RadioQA and ExtendedQA datasets. Agentic retrieval significantly improved mean diagnostic accuracy over zero-shot prompting (73% vs. 64%; P<0.001) and conventional online RAG (73% vs. 68%; P<0.001). The greatest gains occurred in mid-sized models (e.g., Mistral Large improved from 72% to 81%) and small-scale models (e.g., Qwen 2.5-7B improved from 55% to 71%), while very large models (>200B parameters) demonstrated minimal changes (<2% improvement). Additionally, agentic retrieval reduced hallucinations (mean 9.4%) and retrieved clinically relevant context in 46% of cases, substantially aiding factual grounding. Even clinically fine-tuned models exhibited meaningful improvements (e.g., MedGemma-27B improved from 71% to 81%), indicating complementary roles of retrieval and fine-tuning. These results highlight the potential of agentic frameworks to enhance factuality and diagnostic accuracy in radiology QA, particularly among mid-sized LLMs, warranting future studies to validate their clinical utility.
The proliferation of Large Language Models (LLMs) in medicine has enabled impressive capabilities, yet a critical gap remains in their ability to perform systematic, transparent, and verifiable reasoning, a cornerstone of clinical practice. This has catalyzed a shift from single-step answer generation to the development of LLMs explicitly designed for medical reasoning. This paper provides the first systematic review of this emerging field. We propose a taxonomy of reasoning enhancement techniques, categorized into training-time strategies (e.g., supervised fine-tuning, reinforcement learning) and test-time mechanisms (e.g., prompt engineering, multi-agent systems). We analyze how these techniques are applied across different data modalities (text, image, code) and in key clinical applications such as diagnosis, education, and treatment planning. Furthermore, we survey the evolution of evaluation benchmarks from simple accuracy metrics to sophisticated assessments of reasoning quality and visual interpretability. Based on an analysis of 60 seminal studies from 2022-2025, we conclude by identifying critical challenges, including the faithfulness-plausibility gap and the need for native multimodal reasoning, and outlining future directions toward building efficient, robust, and sociotechnically responsible medical AI.
While AI excels at generating text, audio, images, and videos, creating interactive audio-visual content such as video games remains challenging. Current LLMs can generate JavaScript games and animations, but lack automated evaluation metrics and struggle with complex content that normally requires teams of humans working for many months (multi-shot, multi-agents) using assets made by artists. To tackle these issues, we built a new metric and a multi-agent system. We propose AVR-Eval, a relative metric for multimedia content quality using Audio-Visual Recordings (AVRs). An omni-modal model (processing text, video, and audio) compares the AVRs of two contents, with a text model reviewing evaluations to determine superiority. We show that AVR-Eval properly identifies good from broken or mismatched content. We built AVR-Agent, a multi-agent system generating JavaScript code from a bank of multimedia assets (audio, images, 3D models). The coding agent selects relevant assets, generates multiple initial codes, uses AVR-Eval to identify the best version, and iteratively improves it through omni-modal agent feedback from the AVR. We run experiments on games and animations with AVR-Eval (win rate of content A against B). We find that content generated by AVR-Agent has a significantly higher win rate against content made through one-shot generation. However, models struggle to leverage custom assets and AVR feedback effectively, showing no higher win rate. This reveals a critical gap: while humans benefit from high-quality assets and audio-visual feedback, current coding models do not seem to utilize these resources as effectively, highlighting fundamental differences between human and machine content creation approaches.
In financial trading, large language model (LLM)-based agents demonstrate significant potential. However, the high sensitivity to market noise undermines the performance of LLM-based trading systems. To address this limitation, we propose a novel multi-agent system featuring an internal competitive mechanism inspired by modern corporate management structures. The system consists of two specialized teams: (1) Data Team - responsible for processing and condensing massive market data into diversified text factors, ensuring they fit the model's constrained context. (2) Research Team - tasked with making parallelized multipath trading decisions based on deep research methods. The core innovation lies in implementing a real-time evaluation and ranking mechanism within each team, driven by authentic market feedback. Each agent's performance undergoes continuous scoring and ranking, with only outputs from top-performing agents being adopted. The design enables the system to adaptively adjust to dynamic environment, enhances robustness against market noise and ultimately delivers superior trading performance. Experimental results demonstrate that our proposed system significantly outperforms prevailing multiagent systems and traditional quantitative investment methods across diverse evaluation metrics.
Large Language Model (LLM) agents exhibit powerful autonomous capabilities across domains such as robotics, virtual assistants, and web automation. However, their stochastic behavior introduces significant safety risks that are difficult to anticipate. Existing rule-based enforcement systems, such as AgentSpec, focus on developing reactive safety rules, which typically respond only when unsafe behavior is imminent or has already occurred. These systems lack foresight and struggle with long-horizon dependencies and distribution shifts. To address these limitations, we propose Pro2Guard, a proactive runtime enforcement framework grounded in probabilistic reachability analysis. Pro2Guard abstracts agent behaviors into symbolic states and learns a Discrete-Time Markov Chain (DTMC) from execution traces. At runtime, it anticipates future risks by estimating the probability of reaching unsafe states, triggering interventions before violations occur when the predicted risk exceeds a user-defined threshold. By incorporating semantic validity checks and leveraging PAC bounds, Pro2Guard ensures statistical reliability while approximating the underlying ground-truth model. We evaluate Pro2Guard extensively across two safety-critical domains: embodied household agents and autonomous vehicles. In embodied agent tasks, Pro2Guard enforces safety early on up to 93.6% of unsafe tasks using low thresholds, while configurable modes (e.g., reflect) allow balancing safety with task success, maintaining up to 80.4% task completion. In autonomous driving scenarios, Pro2Guard achieves 100% prediction of traffic law violations and collisions, anticipating risks up to 38.66 seconds ahead.
Modern cyber attacks unfold through multiple stages, requiring defenders to dynamically prioritize mitigations under uncertainty. While game-theoretic models capture attacker-defender interactions, existing approaches often rely on static assumptions and lack integration with real-time threat intelligence, limiting their adaptability. This paper presents CyGATE, a game-theoretic framework modeling attacker-defender interactions, using large language models (LLMs) with retrieval-augmented generation (RAG) to enhance tactic selection and patch prioritization. Applied to a two-agent scenario, CyGATE frames cyber conflicts as a partially observable stochastic game (POSG) across Cyber Kill Chain stages. Both agents use belief states to navigate uncertainty, with the attacker adapting tactics and the defender re-prioritizing patches based on evolving risks and observed adversary behavior. The framework's flexible architecture enables extension to multi-agent scenarios involving coordinated attackers, collaborative defenders, or complex enterprise environments with multiple stakeholders. Evaluated in a dynamic patch scheduling scenario, CyGATE effectively prioritizes high-risk vulnerabilities, enhancing adaptability through dynamic threat integration, strategic foresight by anticipating attacker moves under uncertainty, and efficiency by optimizing resource use.
Graph Neural Networks (GNNs) have achieved remarkable success in graph-based learning by propagating information among neighbor nodes via predefined aggregation mechanisms. However, such fixed schemes often suffer from two key limitations. First, they cannot handle the imbalance in node informativeness -- some nodes are rich in information, while others remain sparse. Second, predefined message passing primarily leverages local structural similarity while ignoring global semantic relationships across the graph, limiting the model's ability to capture distant but relevant information. We propose Retrieval-augmented Graph Agentic Network (ReaGAN), an agent-based framework that empowers each node with autonomous, node-level decision-making. Each node acts as an agent that independently plans its next action based on its internal memory, enabling node-level planning and adaptive message propagation. Additionally, retrieval-augmented generation (RAG) allows nodes to access semantically relevant content and build global relationships in the graph. ReaGAN achieves competitive performance under few-shot in-context settings using a frozen LLM backbone without fine-tuning, showcasing the potential of agentic planning and local-global retrieval in graph learning.
Large Language Models (LLMs) have shown remarkable advancements in tackling agent-oriented tasks. Despite their potential, existing work faces challenges when deploying LLMs in agent-based environments. The widely adopted agent paradigm ReAct centers on integrating single-step reasoning with immediate action execution, which limits its effectiveness in complex tasks requiring long-term strategic planning. Furthermore, the coordination between the planner and executor during problem-solving is also a critical factor to consider in agent design. Additionally, current approaches predominantly rely on supervised fine-tuning, which often leads models to memorize established task completion trajectories, thereby restricting their generalization ability when confronted with novel problem contexts. To address these challenges, we introduce an adaptive global plan-based agent paradigm AdaPlan, aiming to synergize high-level explicit guidance with execution to support effective long-horizon decision-making. Based on the proposed paradigm, we further put forward PilotRL, a global planning-guided training framework for LLM agents driven by progressive reinforcement learning. We first develop the model's ability to follow explicit guidance from global plans when addressing agent tasks. Subsequently, based on this foundation, we focus on optimizing the quality of generated plans. Finally, we conduct joint optimization of the model's planning and execution coordination. Experiments indicate that PilotRL could achieve state-of-the-art performances, with LLaMA3.1-8B-Instruct + PilotRL surpassing closed-sourced GPT-4o by 3.60%, while showing a more substantial gain of 55.78% comparing to GPT-4o-mini at a comparable parameter scale.
Humans constantly generate a diverse range of tasks guided by internal motivations. While generative agents powered by large language models (LLMs) aim to simulate this complex behavior, it remains uncertain whether they operate on similar cognitive principles. To address this, we conducted a task-generation experiment comparing human responses with those of an LLM agent (GPT-4o). We find that human task generation is consistently influenced by psychological drivers, including personal values (e.g., Openness to Change) and cognitive style. Even when these psychological drivers are explicitly provided to the LLM, it fails to reflect the corresponding behavioral patterns. They produce tasks that are markedly less social, less physical, and thematically biased toward abstraction. Interestingly, while the LLM's tasks were perceived as more fun and novel, this highlights a disconnect between its linguistic proficiency and its capacity to generate human-like, embodied goals.We conclude that there is a core gap between the value-driven, embodied nature of human cognition and the statistical patterns of LLMs, highlighting the necessity of incorporating intrinsic motivation and physical grounding into the design of more human-aligned agents.
Recent advances in natural language processing (NLP) have opened up greater opportunities to enable fine-tuned large language models (LLMs) to behave as more powerful interactive agents through improved instruction-following ability. However, understanding how this impacts confidence calibration for reliable model output has not been researched in full. In this work, we examine various open-sourced LLMs, identifying significant calibration degradation after instruction tuning in each. Seeking a practical solution, we look towards label smoothing, which has been shown as an effective method to regularize for overconfident predictions but has yet to be widely adopted in the supervised fine-tuning (SFT) of LLMs. We first provide insight as to why label smoothing is sufficient to maintain calibration throughout the SFT process. However, settings remain where the effectiveness of smoothing is severely diminished, in particular the case of large vocabulary LLMs (LV-LLMs). We posit the cause to stem from the ability to become over-confident, which has a direct relationship with the hidden size and vocabulary size, and justify this theoretically and experimentally. Finally, we address an outstanding issue regarding the memory footprint of the cross-entropy loss computation in the label smoothed loss setting, designing a customized kernel to dramatically reduce memory consumption without sacrificing speed or performance in comparison to existing solutions for non-smoothed losses.
Large Language Models (LLMs) have demonstrated remarkable capabilities, leading to a significant increase in user demand for LLM services. However, cloud-based LLM services often suffer from high latency, unstable responsiveness, and privacy concerns. Therefore, multiple LLMs are usually deployed at the network edge to boost real-time responsiveness and protect data privacy, particularly for many emerging smart mobile and IoT applications. Given the varying response quality and latency of LLM services, a critical issue is how to route user requests from mobile and IoT devices to an appropriate LLM service (i.e., edge LLM expert) to ensure acceptable quality-of-service (QoS). Existing routing algorithms fail to simultaneously address the heterogeneity of LLM services, the interference among requests, and the dynamic workloads necessary for maintaining long-term stable QoS. To meet these challenges, in this paper we propose a novel deep reinforcement learning (DRL)-based QoS-aware LLM routing framework for sustained high-quality LLM services. Due to the dynamic nature of the global state, we propose a dynamic state abstraction technique to compactly represent global state features with a heterogeneous graph attention network (HAN). Additionally, we introduce an action impact estimator and a tailored reward function to guide the DRL agent in maximizing QoS and preventing latency violations. Extensive experiments on both Poisson and real-world workloads demonstrate that our proposed algorithm significantly improves average QoS and computing resource efficiency compared to existing baselines.
Code generation agents powered by large language models (LLMs) are revolutionizing the software development paradigm. Distinct from previous code generation techniques, code generation agents are characterized by three core features. 1) Autonomy: the ability to independently manage the entire workflow, from task decomposition to coding and debugging. 2) Expanded task scope: capabilities that extend beyond generating code snippets to encompass the full software development lifecycle (SDLC). 3) Enhancement of engineering practicality: a shift in research emphasis from algorithmic innovation toward practical engineering challenges, such as system reliability, process management, and tool integration. This domain has recently witnessed rapid development and an explosion in research, demonstrating significant application potential. This paper presents a systematic survey of the field of LLM-based code generation agents. We trace the technology's developmental trajectory from its inception and systematically categorize its core techniques, including both single-agent and multi-agent architectures. Furthermore, this survey details the applications of LLM-based agents across the full SDLC, summarizes mainstream evaluation benchmarks and metrics, and catalogs representative tools. Finally, by analyzing the primary challenges, we identify and propose several foundational, long-term research directions for the future work of the field.
The discipline of physics stands as a cornerstone of human intellect, driving the evolution of technology and deepening our understanding of the fundamental principles of the cosmos. Contemporary literature includes some works centered on the task of solving physics problems - a crucial domain of natural language reasoning. In this paper, we evaluate the performance of frontier LLMs in solving physics problems, both mathematical and descriptive. We also employ a plethora of inference-time techniques and agentic frameworks to improve the performance of the models. This includes the verification of proposed solutions in a cumulative fashion by other, smaller LLM agents, and we perform a comparative analysis of the performance that the techniques entail. There are significant improvements when the multi-agent framework is applied to problems that the models initially perform poorly on. Furthermore, we introduce a new evaluation benchmark for physics problems, ${\rm P{\small HYSICS}E{\small VAL}}$, consisting of 19,609 problems sourced from various physics textbooks and their corresponding correct solutions scraped from physics forums and educational websites. Our code and data are publicly available at https://github.com/areebuzair/PhysicsEval.
AI agents built on large language models (LLMs) hold enormous promise, but current practice focuses on a one-task-one-agent approach, which not only falls short of scalability and generality, but also suffers from the fundamental limitations of autoregressive LLMs. On the other hand, humans are general agents who reason by mentally simulating the outcomes of their actions and plans. Moving towards a more general and powerful AI agent, we introduce SimuRA, a goal-oriented architecture for generalized agentic reasoning. Based on a principled formulation of optimal agent in any environment, \modelname overcomes the limitations of autoregressive reasoning by introducing a world model for planning via simulation. The generalized world model is implemented using LLM, which can flexibly plan in a wide range of environments using the concept-rich latent space of natural language. Experiments on difficult web browsing tasks show that \modelname improves the success of flight search from 0\% to 32.2\%. World-model-based planning, in particular, shows consistent advantage of up to 124\% over autoregressive planning, demonstrating the advantage of world model simulation as a reasoning paradigm. We are excited about the possibility for training a single, general agent model based on LLMs that can act superintelligently in all environments. To start, we make SimuRA, a web-browsing agent built on \modelname with pretrained LLMs, available as a research demo for public testing.
Evaluating AI agents within complex, interactive environments that mirror real-world challenges is critical for understanding their practical capabilities. While existing agent benchmarks effectively assess skills like tool use or performance on structured tasks, they often do not fully capture an agent's ability to operate autonomously in exploratory environments that demand sustained, self-directed reasoning over a long and growing context. To spur the development of agents capable of more robust intrinsic reasoning over long horizons, we introduce TextQuests, a benchmark based on the Infocom suite of interactive fiction games. These text-based adventures, which can take human players over 30 hours and require hundreds of precise actions to solve, serve as an effective proxy for evaluating AI agents on focused, stateful tasks. The benchmark is specifically designed to assess an LLM agent's capacity for self-contained problem-solving by precluding the use of external tools, thereby focusing on intrinsic long-context reasoning capabilities in an exploratory environment characterized by the need for trial-and-error learning and sustained problem-solving within a single interactive session. We release TextQuests at https://textquests.ai.
We provide a comprehensive examination of agent-based approaches that codify the principles and linkages underlying multi-agent systems, simulations, and information systems. Based on two decades of study, this paper confirms a framework intended as a formal specification for geosimulation platforms. Our findings show that large language models (LLMs) can be effectively incorporated as agent components if they follow a structured architecture specific to fundamental agent activities such as perception, memory, planning, and action. This integration is precisely consistent with the architecture that we formalize, providing a solid platform for next-generation geosimulation systems.
We introduce CFDagent, a zero-shot, multi-agent system that enables fully autonomous computational fluid dynamics (CFD) simulations from natural language prompts. CFDagent integrates three specialized LLM-driven agents: (i) the Preprocessing Agent that generates 3D geometries from textual or visual inputs using a hybrid text-to-3D diffusion model (Point-E) and automatically meshes the geometries; (ii) the Solver Agent that configures and executes an immersed boundary flow solver; and (iii) the Postprocessing Agent that analyzes and visualizes the results, including multimodal renderings. These agents are interactively guided by GPT-4o via conversational prompts, enabling intuitive and user-friendly interaction. We validate CFDagent by reproducing canonical sphere flows at Reynolds numbers of 100 and 300 using three distinct inputs: a simple text prompt (i.e., "sphere"), an image-based input, and a standard sphere model. The computed drag and lift coefficients from meshes produced by each input approach closely match available data. The proposed system enables synthesization of flow simulations and photorealistic visualizations for complex geometries. Through extensive tests on canonical and realistic scenarios, we demonstrate the robustness, versatility, and practical applicability of CFDagent. By bridging generative AI with high-fidelity simulations, CFDagent significantly lowers barriers to expert-level CFD, unlocking broad opportunities in education, scientific research, and practical engineering applications.
Large Language Models (LLMs) process millions of queries daily, making efficient response caching a compelling optimization for reducing cost and latency. However, preserving relevance to user queries using this approach proves difficult due to the personalized nature of chatbot interactions and the limited accuracy of semantic similarity search. To address this, we present TweakLLM, a novel routing architecture that employs a lightweight LLM to dynamically adapt cached responses to incoming prompts. Through comprehensive evaluation, including user studies with side-by-side comparisons, satisfaction voting, as well as multi-agent LLM debates, we demonstrate that TweakLLM maintains response quality comparable to frontier models while significantly improving cache effectiveness. Our results across real-world datasets highlight TweakLLM as a scalable, resource-efficient caching solution for high-volume LLM deployments without compromising user experience.
Agent-assisted memory recall is one critical research problem in the field of human-computer interaction. In conventional methods, the agent can retrieve information from its equipped memory module to help the person recall incomplete or vague memories. The limited size of memory module hinders the acquisition of complete memories and impacts the memory recall performance in practice. Memory theories suggest that the person's relevant memory can be proactively activated through some effective cues. Inspired by this, we propose a novel strategy-guided agent-assisted memory recall method, allowing the agent to transform an original query into a cue-rich one via the judiciously designed strategy to help the person recall memories. To this end, there are two key challenges. (1) How to choose the appropriate recall strategy for diverse forgetting scenarios with distinct memory-recall characteristics? (2) How to obtain the high-quality responses leveraging recall strategies, given only abstract and sparsely annotated strategy patterns? To address the challenges, we propose a Recall Router framework. Specifically, we design a 5W Recall Map to classify memory queries into five typical scenarios and define fifteen recall strategy patterns across the corresponding scenarios. We then propose a hierarchical recall tree combined with the Monte Carlo Tree Search algorithm to optimize the selection of strategy and the generation of strategy responses. We construct an instruction tuning dataset and fine-tune multiple open-source large language models (LLMs) to develop MemoCue, an agent that excels in providing memory-inspired responses. Experiments on three representative datasets show that MemoCue surpasses LLM-based methods by 17.74% in recall inspiration. Further human evaluation highlights its advantages in memory-recall applications.
Large language model-based agents, empowered by in-context learning (ICL), have demonstrated strong capabilities in complex reasoning and tool-use tasks. However, existing works have shown that the effectiveness of ICL is highly sensitive to the choice of demonstrations, with suboptimal examples often leading to unstable or degraded performance. While prior work has explored example selection, including in some agentic or multi-step settings, existing approaches typically rely on heuristics or task-specific designs and lack a general, theoretically grounded criterion for what constitutes an effective demonstration across reasoning steps. Therefore, it is non-trivial to develop a principled, general-purpose method for selecting demonstrations that consistently benefit agent performance. In this paper, we address this challenge with DICE, Dynamic In-Context Example Selection for LLM Agents, a theoretically grounded ICL framework for agentic tasks that selects the most relevant demonstrations at each step of reasoning. Our approach decomposes demonstration knowledge into transferable and non-transferable components through a causal lens, showing how the latter can introduce spurious dependencies that impair generalization. We further propose a stepwise selection criterion with a formal guarantee of improved agent performance. Importantly, DICE is a general, framework-agnostic solution that can be integrated as a plug-in module into existing agentic frameworks without any additional training cost. Extensive experiments across diverse domains demonstrate our method's effectiveness and generality, highlighting the importance of principled, context-aware demo selection for robust and efficient LLM agents.
This paper presents the design, implementation, and evaluation behind a Large Language Model (LLM) agent that chats with an industrial production-grade ERP system. The agent is capable of interpreting natural language queries and translating them into executable SQL statements, leveraging open-weight LLMs. A novel dual-agent architecture combining reasoning and critique stages was proposed to improve query generation reliability.
Software issue resolution is a critical challenge in software engineering and has garnered increasing attention in recent years. With the rapid advancement of large language models (LLMs), substantial progress has been made in addressing real-world software engineering tasks. Recent studies have introduced ensemble reasoning techniques to enhance the performance of LLM-based issue resolution. However, existing prompting-based methods still face limitations in effectively exploring large ensemble spaces and lack the capacity for repository-level understanding, both of which constrain their overall effectiveness. In this paper, we propose Trae Agent, the first agent-based ensemble reasoning approach for repository-level issue resolution. Trae Agent formulates our goal as an optimal solution search problem and addresses two key challenges, i.e., large ensemble spaces and repository-level understanding, through modular agents for generation, pruning, and selection. We conduct extensive experiments using three leading LLMs on the widely-adopted SWE-bench benchmark, comparing Trae Agent against four state-of-the-art ensemble reasoning techniques. Experimental results demonstrate that Trae Agent consistently achieves superior performance, with an average improvement of 10.22% over all baselines in terms of Pass@1. Trae Agent has achieved first place on the SWE-bench Verified leaderboard, with a notable Pass@1 score of 75.20%. We are pleased to release Trae Agent as an open-source project to support the research community, with all resources available at https://github.com/bytedance/trae-agent.
Recent advances in large language model (LLM) agents have shown remarkable progress in software issue resolution, leveraging advanced techniques such as multi-agent collaboration and Monte Carlo Tree Search (MCTS). However, current agents act as memoryless explorers - treating each problem separately without retaining or reusing knowledge from previous repair experiences. This leads to redundant exploration of failed trajectories and missed chances to adapt successful issue resolution methods to similar problems. To address this problem, we introduce SWE-Exp, an experience - enhanced approach that distills concise and actionable experience from prior agent trajectories, enabling continuous learning across issues. Our method introduces a multi-faceted experience bank that captures both successful and failed repair attempts. Specifically, it extracts reusable issue resolution knowledge at different levels - from high-level problem comprehension to specific code changes. Experiments show that SWE-Exp achieves state-of-the-art resolution rate (41.6% Pass@1) on SWE-bench-Verified under open-source agent frameworks. Our approach establishes a new paradigm in which automated software engineering agents systematically accumulate and leverage repair expertise, fundamentally shifting from trial-and-error exploration to strategic, experience-driven issue resolution.
Issue resolution has made remarkable progress thanks to the advanced reasoning capabilities of large language models (LLMs). Recently, agent-based frameworks such as SWE-agent have further advanced this progress by enabling autonomous, tool-using agents to tackle complex software engineering tasks. While existing agent-based issue resolution approaches are primarily based on agents' independent explorations, they often get stuck in local solutions and fail to identify issue patterns that span across different parts of the codebase. To address this limitation, we propose SWE-Debate, a competitive multi-agent debate framework that encourages diverse reasoning paths and achieves more consolidated issue localization. SWE-Debate first creates multiple fault propagation traces as localization proposals by traversing a code dependency graph. Then, it organizes a three-round debate among specialized agents, each embodying distinct reasoning perspectives along the fault propagation trace. This structured competition enables agents to collaboratively converge on a consolidated fix plan. Finally, this consolidated fix plan is integrated into an MCTS-based code modification agent for patch generation. Experiments on the SWE-bench benchmark show that SWE-Debate achieves new state-of-the-art results in open-source agent frameworks and outperforms baselines by a large margin.
Recent advances in large language models (LLMs) have significantly impacted data science workflows, giving rise to specialized data science agents designed to automate analytical tasks. Despite rapid adoption, systematic benchmarks evaluating the efficacy and limitations of these agents remain scarce. In this paper, we introduce a comprehensive benchmark specifically crafted to reflect real-world user interactions with data science agents by observing usage of our commercial applications. We evaluate three LLMs: Claude-4.0-Sonnet, Gemini-2.5-Flash, and OpenAI-o4-Mini across three approaches: zero-shot with context engineering, multi-step with context engineering, and with SmolAgent. Our benchmark assesses performance across a diverse set of eight data science task categories, additionally exploring the sensitivity of models to common prompting issues, such as data leakage and slightly ambiguous instructions. We further investigate the influence of temperature parameters on overall and task-specific outcomes for each model and approach. Our findings reveal distinct performance disparities among the evaluated models and methodologies, highlighting critical factors that affect practical deployment. The benchmark dataset and evaluation framework introduced herein aim to provide a foundation for future research of more robust and effective data science agents.
The emergence of large language models (LLMs) is propelling automated scientific discovery to the next level, with LLM-based Artificial Intelligence (AI) Scientist systems now taking the lead in scientific research. Several influential works have already appeared in the field of AI Scientist systems, with AI-generated research papers having been accepted at the ICLR 2025 workshop, suggesting that a human-level AI Scientist capable of uncovering phenomena previously unknown to humans, may soon become a reality. In this survey, we focus on the central question: How far are AI scientists from changing the world and reshaping the scientific research paradigm? To answer this question, we provide a prospect-driven review that comprehensively analyzes the current achievements of AI Scientist systems, identifying key bottlenecks and the critical components required for the emergence of a scientific agent capable of producing ground-breaking discoveries that solve grand challenges. We hope this survey will contribute to a clearer understanding of limitations of current AI Scientist systems, showing where we are, what is missing, and what the ultimate goals for scientific AI should be.
Current multi-agent systems (MAS) frameworks often rely on manually designed and static collaboration graph structures, limiting adaptability and performance. To address these limitations, we propose DynaSwarm, a dynamic framework that enhances LLM-based MAS through two key innovations: (1) an actor-critic reinforcement learning (A2C) mechanism to optimize graph structures with improved stability over prior RL methods, and (2) a dynamic graph selector that adaptively chooses the optimal graph structure for each input sample via parameter-efficient LLM fine-tuning. DynaSwarm eliminates the need for rigid, one-fits-all graph architectures, instead leveraging sample-specific idiosyncrasies to dynamically route queries through specialized agent networks. (c) We propose to fine-tune the demonstration retriever to fully exploit the power of in-context learning (ICL). Extensive experiments on question answering, mathematical reasoning, and coding tasks demonstrate that DynaSwarm consistently outperforms state-of-the-art single-agent and MAS baselines across multiple LLM backbones. Our findings highlight the importance of sample-aware structural flexibility in LLM MAS designs.
Early and accurate diagnosis of Alzheimer's disease (AD), a complex neurodegenerative disorder, requires analysis of heterogeneous biomarkers (e.g., neuroimaging, genetic risk factors, cognitive tests, and cerebrospinal fluid proteins) typically represented in a tabular format. With flexible few-shot reasoning, multimodal integration, and natural-language-based interpretability, large language models (LLMs) offer unprecedented opportunities for prediction with structured biomedical data. We propose a novel framework called TAP-GPT, Tabular Alzheimer's Prediction GPT, that adapts TableGPT2, a multimodal tabular-specialized LLM originally developed for business intelligence tasks, for AD diagnosis using structured biomarker data with small sample sizes. Our approach constructs few-shot tabular prompts using in-context learning examples from structured biomedical data and finetunes TableGPT2 using the parameter-efficient qLoRA adaption for a clinical binary classification task of AD or cognitively normal (CN). The TAP-GPT framework harnesses the powerful tabular understanding ability of TableGPT2 and the encoded prior knowledge of LLMs to outperform more advanced general-purpose LLMs and a tabular foundation model (TFM) developed for prediction tasks. To our knowledge, this is the first application of LLMs to the prediction task using tabular biomarker data, paving the way for future LLM-driven multi-agent frameworks in biomedical informatics.
The demand for AI-generated GPU kernels is rapidly growing, influenced by the need for scalable, hardware-optimized solutions in both industry and academia. As deep learning workloads grow in complexity and diversity, it is imperative to automate low-level kernel development to meet performance and productivity demands. Major cloud providers, semiconductor companies, and research institutions are now investing heavily in AI-driven code generation for GPUs, aiming to reduce manual optimization efforts while achieving near-expert performance on hardware like AMD MI300X. The Triton language, a Python-based DSL for GPU programming, has emerged as a popular target for such AI-generated kernels due to its balance of performance and ease-of-coding. In this work, we present an evaluation suite for Triton-based GPU kernels and GEAK (Generating Efficient AI-centric GPU Kernels)-a framework that leverages cutting-edge LLMs to generate performant Triton code specifically for AMD GPUs, including the AMD MI300X and MI250. GEAK leverages inference-time compute scaling to produce Triton-based GPU kernels using a reasoning loop adapted from Reflexion-style feedback mechanisms. On two evaluation benchmarks, GEAK significantly outperformed the baselines of directly prompting frontier LLMs as well as Reflexion-based generation pipelines by achieving correctness up to $63$% and execution speed up of up to $2.59$X. These results highlight the promise of GEAK-like agentic code generation for accelerating the adoption of diverse hardware platforms and democratizing access to expert-level kernel performance.
In this work, we study a critical research problem regarding the trustworthiness of large language models (LLMs): how LLMs behave when encountering ambiguous narrative text, with a particular focus on Chinese textual ambiguity. We created a benchmark dataset by collecting and generating ambiguous sentences with context and their corresponding disambiguated pairs, representing multiple possible interpretations. These annotated examples are systematically categorized into 3 main categories and 9 subcategories. Through experiments, we discovered significant fragility in LLMs when handling ambiguity, revealing behavior that differs substantially from humans. Specifically, LLMs cannot reliably distinguish ambiguous text from unambiguous text, show overconfidence in interpreting ambiguous text as having a single meaning rather than multiple meanings, and exhibit overthinking when attempting to understand the various possible meanings. Our findings highlight a fundamental limitation in current LLMs that has significant implications for their deployment in real-world applications where linguistic ambiguity is common, calling for improved approaches to handle uncertainty in language understanding. The dataset and code are publicly available at this GitHub repository: https://github.com/ictup/LLM-Chinese-Textual-Disambiguation.