Synthesizing interactive 3D scenes from text is essential for gaming, virtual reality, and embodied AI. However, existing methods face several challenges. Learning-based approaches depend on small-scale indoor datasets, limiting the scene diversity and layout complexity. While large language models (LLMs) can leverage diverse text-domain knowledge, they struggle with spatial realism, often producing unnatural object placements that fail to respect common sense. Our key insight is that vision perception can bridge this gap by providing realistic spatial guidance that LLMs lack. To this end, we introduce Scenethesis, a training-free agentic framework that integrates LLM-based scene planning with vision-guided layout refinement. Given a text prompt, Scenethesis first employs an LLM to draft a coarse layout. A vision module then refines it by generating an image guidance and extracting scene structure to capture inter-object relations. Next, an optimization module iteratively enforces accurate pose alignment and physical plausibility, preventing artifacts like object penetration and instability. Finally, a judge module verifies spatial coherence. Comprehensive experiments show that Scenethesis generates diverse, realistic, and physically plausible 3D interactive scenes, making it valuable for virtual content creation, simulation environments, and embodied AI research.
Recent advancements have significantly enhanced the performance of large language models (LLMs) in tackling complex reasoning tasks, achieving notable success in domains like mathematical and logical reasoning. However, these methods encounter challenges with complex planning tasks, primarily due to extended reasoning steps, diverse constraints, and the challenge of handling multiple distinct sub-tasks. To address these challenges, we propose HyperTree Planning (HTP), a novel reasoning paradigm that constructs hypertree-structured planning outlines for effective planning. The hypertree structure enables LLMs to engage in hierarchical thinking by flexibly employing the divide-and-conquer strategy, effectively breaking down intricate reasoning steps, accommodating diverse constraints, and managing multiple distinct sub-tasks in a well-organized manner. We further introduce an autonomous planning framework that completes the planning process by iteratively refining and expanding the hypertree-structured planning outlines. Experiments demonstrate the effectiveness of HTP, achieving state-of-the-art accuracy on the TravelPlanner benchmark with Gemini-1.5-Pro, resulting in a 3.6 times performance improvement over o1-preview.
Partially Observable Markov Decision Processes (POMDPs) model decision making under uncertainty. While there are many approaches to approximately solving POMDPs, we aim to address the problem of learning such models. In particular, we are interested in a subclass of POMDPs wherein the components of the model, including the observation function, reward function, transition function, and initial state distribution function, can be modeled as low-complexity probabilistic graphical models in the form of a short probabilistic program. Our strategy to learn these programs uses an LLM as a prior, generating candidate probabilistic programs that are then tested against the empirical distribution and adjusted through feedback. We experiment on a number of classical toy POMDP problems, simulated MiniGrid domains, and two real mobile-base robotics search domains involving partial observability. Our results show that using an LLM to guide in the construction of a low-complexity POMDP model can be more effective than tabular POMDP learning, behavior cloning, or direct LLM planning.
Immersive Virtual Reality (iVR) applications have shown immense potential for skill training and learning in manufacturing. However, authoring of such applications requires technical expertise, which makes it difficult for educators to author instructions targeted at desired learning outcomes. We present FlowTrainer, an LLM-assisted interactive system to allow educators to author lesson plans for their iVR instruction based on desired goals. The authoring workflow is supported by Backward design to align the planned lesson based on the desired outcomes. We implemented a welding use case and conducted a user study with welding experts to test the effectiveness of the system in authoring outcome-oriented lesson plans. The study results showed that the system allowed users to plan lesson plans based on desired outcomes while reducing the time and technical expertise required for the authoring process. We believe that such efforts can allow widespread adoption of iVR solutions in manufacturing training to meet the workforce demands in the industry.
Large language models (LLMs) have demonstrated remarkable capabilities in natural language understanding and task generalization. However, their application to structured data analysis remains fragile due to inconsistencies in schema interpretation, misalignment between user intent and model output, and limited mechanisms for self-correction when failures occur. This paper introduces the STROT Framework (Structured Task Reasoning and Output Transformation), a method for structured prompting and feedback-driven transformation logic generation aimed at improving the reliability and semantic alignment of LLM-based analytical workflows. STROT begins with lightweight schema introspection and sample-based field classification, enabling dynamic context construction that captures both the structure and statistical profile of the input data. This contextual information is embedded in structured prompts that guide the model toward generating task-specific, interpretable outputs. To address common failure modes in complex queries, STROT incorporates a refinement mechanism in which the model iteratively revises its outputs based on execution feedback and validation signals. Unlike conventional approaches that rely on static prompts or single-shot inference, STROT treats the LLM as a reasoning agent embedded within a controlled analysis loop -- capable of adjusting its output trajectory through planning and correction. The result is a robust and reproducible framework for reasoning over structured data with LLMs, applicable to diverse data exploration and analysis tasks where interpretability, stability, and correctness are essential.
The growing capabilities of large language models (LLMs) in instruction-following and context-understanding lead to the era of agents with numerous applications. Among these, task planning agents have become especially prominent in realistic scenarios involving complex internal pipelines, such as context understanding, tool management, and response generation. However, existing benchmarks predominantly evaluate agent performance based on task completion as a proxy for overall effectiveness. We hypothesize that merely improving task completion is misaligned with maximizing user satisfaction, as users interact with the entire agentic process and not only the end result. To address this gap, we propose PIPA, a unified evaluation protocol that conceptualizes the behavioral process of interactive task planning agents within a partially observable Markov Decision Process (POMDP) paradigm. The proposed protocol offers a comprehensive assessment of agent performance through a set of atomic evaluation criteria, allowing researchers and practitioners to diagnose specific strengths and weaknesses within the agent's decision-making pipeline. Our analyses show that agents excel in different behavioral stages, with user satisfaction shaped by both outcomes and intermediate behaviors. We also highlight future directions, including systems that leverage multiple agents and the limitations of user simulators in task planning.
The rapid evolution of wireless networks presents unprecedented challenges in managing complex and dynamic systems. Existing methods are increasingly facing fundamental limitations in addressing these challenges. In this paper, we introduce WirelessAgent, a novel framework that harnesses large language models (LLMs) to create autonomous AI agents for diverse wireless network tasks. This framework integrates four core modules that mirror human cognitive processes: perception, memory, planning, and action. To implement it, we provide a basic usage based on agentic workflows and the LangGraph architecture. We demonstrate the effectiveness of WirelessAgent through a comprehensive case study on network slicing. The numerical results show that WirelessAgent achieves $44.4\%$ higher bandwidth utilization than the \emph{Prompt-based} method, while performing only $4.3\%$ below the \emph{Rule-based optimality}. Notably, WirelessAgent delivers near-optimal network throughput across diverse network scenarios. These underscore the framework's potential for intelligent and autonomous resource management in future wireless networks. The code is available at \url{https://github.com/jwentong/WirelessAgent_R1}.
Planning is essential for artificial intelligence systems to look ahead and proactively determine a course of actions to reach objectives in the virtual and real world. Recent work on large language models (LLMs) sheds light on their planning capability in various tasks. However, it remains unclear what signals in the context influence the model performance. In this work, we explore how to improve the model planning capability through in-context learning (ICL), specifically, what signals can help select the exemplars. Through extensive experiments, we observe that commonly used problem similarity may result in false positives with drastically different plans, which can mislead the model. In response, we propose to sample and filter exemplars leveraging plan side action sequence similarity (AS). We propose GRASE-DC: a two-stage pipeline that first re-samples high AS exemplars and then curates the selected exemplars with dynamic clustering on AS to achieve a balance of relevance and diversity. Our experimental result confirms that GRASE-DC achieves significant performance improvement on various planning tasks (up to ~11-40 point absolute accuracy improvement with 27.3% fewer exemplars needed on average). With GRASE-DC* + VAL, where we iteratively apply GRASE-DC with a validator, we are able to even boost the performance by 18.9% more. Extensive analysis validates the consistent performance improvement of GRASE-DC with various backbone LLMs and on both classical planning and natural language planning benchmarks. GRASE-DC can further boost the planning accuracy by ~24 absolute points on harder problems using simpler problems as exemplars over a random baseline. This demonstrates its ability to generalize to out-of-distribution problems.
Efficient path planning in robotics, particularly within large-scale, dynamic environments, remains a significant hurdle. While Large Language Models (LLMs) offer strong reasoning capabilities, their high computational cost and limited adaptability in dynamic scenarios hinder real-time deployment on edge devices. We present SmallPlan -- a novel framework leveraging LLMs as teacher models to train lightweight Small Language Models (SLMs) for high-level path planning tasks. In SmallPlan, the SLMs provide optimal action sequences to navigate across scene graphs that compactly represent full-scaled 3D scenes. The SLMs are trained in a simulation-powered, interleaved manner with LLM-guided supervised fine-tuning (SFT) and reinforcement learning (RL). This strategy not only enables SLMs to successfully complete navigation tasks but also makes them aware of important factors like travel distance and number of trials. Through experiments, we demonstrate that the fine-tuned SLMs perform competitively with larger models like GPT-4o on sequential path planning, without suffering from hallucination and overfitting. SmallPlan is resource-efficient, making it well-suited for edge-device deployment and advancing practical autonomous robotics.
Rapid advancements in artificial intelligence (AI) have enabled robots to performcomplex tasks autonomously with increasing precision. However, multi-robot systems (MRSs) face challenges in generalization, heterogeneity, and safety, especially when scaling to large-scale deployments like disaster response. Traditional approaches often lack generalization, requiring extensive engineering for new tasks and scenarios, and struggle with managing diverse robots. To overcome these limitations, we propose a Human-in-the-loop Multi-Robot Collaboration Framework (HMCF) powered by large language models (LLMs). LLMs enhance adaptability by reasoning over diverse tasks and robot capabilities, while human oversight ensures safety and reliability, intervening only when necessary. Our framework seamlessly integrates human oversight, LLM agents, and heterogeneous robots to optimize task allocation and execution. Each robot is equipped with an LLM agent capable of understanding its capabilities, converting tasks into executable instructions, and reducing hallucinations through task verification and human supervision. Simulation results show that our framework outperforms state-of-the-art task planning methods, achieving higher task success rates with an improvement of 4.76%. Real-world tests demonstrate its robust zero-shot generalization feature and ability to handle diverse tasks and environments with minimal human intervention.
In response to the lack of trust in Artificial Intelligence (AI) for sequential planning, we design a Computational Tree Logic-guided large language model (LLM)-based natural language explanation framework designed for the Monte Carlo Tree Search (MCTS) algorithm. MCTS is often considered challenging to interpret due to the complexity of its search trees, but our framework is flexible enough to handle a wide range of free-form post-hoc queries and knowledge-based inquiries centered around MCTS and the Markov Decision Process (MDP) of the application domain. By transforming user queries into logic and variable statements, our framework ensures that the evidence obtained from the search tree remains factually consistent with the underlying environmental dynamics and any constraints in the actual stochastic control process. We evaluate the framework rigorously through quantitative assessments, where it demonstrates strong performance in terms of accuracy and factual consistency.
Urban Air Mobility (UAM) is an emerging System of System (SoS) that faces challenges in system architecture, planning, task management, and execution. Traditional architectural approaches struggle with scalability, adaptability, and seamless resource integration within dynamic and complex environments. This paper presents an intelligent holonic architecture that incorporates Large Language Model (LLM) to manage the complexities of UAM. Holons function semi autonomously, allowing for real time coordination among air taxis, ground transport, and vertiports. LLMs process natural language inputs, generate adaptive plans, and manage disruptions such as weather changes or airspace closures.Through a case study of multimodal transportation with electric scooters and air taxis, we demonstrate how this architecture enables dynamic resource allocation, real time replanning, and autonomous adaptation without centralized control, creating more resilient and efficient urban transportation networks. By advancing decentralized control and AI driven adaptability, this work lays the groundwork for resilient, human centric UAM ecosystems, with future efforts targeting hybrid AI integration and real world validation.
With the increasing demand for heterogeneous Unmanned Aerial Vehicle (UAV) swarms to perform complex tasks in urban environments, system design now faces major challenges, including efficient semantic understanding, flexible task planning, and the ability to dynamically adjust coordination strategies in response to evolving environmental conditions and continuously changing task requirements. To address the limitations of existing approaches, this paper proposes coordination field agentic system for coordinating heterogeneous UAV swarms in complex urban scenarios. In this system, large language models (LLMs) is responsible for interpreting high-level human instructions and converting them into executable commands for the UAV swarms, such as patrol and target tracking. Subsequently, a Coordination field mechanism is proposed to guide UAV motion and task selection, enabling decentralized and adaptive allocation of emergent tasks. A total of 50 rounds of comparative testing were conducted across different models in a 2D simulation space to evaluate their performance. Experimental results demonstrate that the proposed system achieves superior performance in terms of task coverage, response time, and adaptability to dynamic changes.
We present an embodied robotic system with an LLM-driven agent-orchestration architecture for autonomous household object management. The system integrates memory-augmented task planning, enabling robots to execute high-level user commands while tracking past actions. It employs three specialized agents: a routing agent, a task planning agent, and a knowledge base agent, each powered by task-specific LLMs. By leveraging in-context learning, our system avoids the need for explicit model training. RAG enables the system to retrieve context from past interactions, enhancing long-term object tracking. A combination of Grounded SAM and LLaMa3.2-Vision provides robust object detection, facilitating semantic scene understanding for task planning. Evaluation across three household scenarios demonstrates high task planning accuracy and an improvement in memory recall due to RAG. Specifically, Qwen2.5 yields best performance for specialized agents, while LLaMA3.1 excels in routing tasks. The source code is available at: https://github.com/marc1198/chat-hsr.
With the rapid advancement of artificial intelligence, there is an increasing demand for intelligent robots capable of assisting humans in daily tasks and performing complex operations. Such robots not only require task planning capabilities but must also execute tasks with stability and robustness. In this paper, we present a closed-loop task planning and acting system, LLM-PAS, which is assisted by a pre-trained Large Language Model (LLM). While LLM-PAS plans long-horizon tasks in a manner similar to traditional task and motion planners, it also emphasizes the execution phase of the task. By transferring part of the constraint-checking process from the planning phase to the execution phase, LLM-PAS enables exploration of the constraint space and delivers more accurate feedback on environmental anomalies during execution. The reasoning capabilities of the LLM allow it to handle anomalies that cannot be addressed by the robust executor. To further enhance the system's ability to assist the planner during replanning, we propose the First Look Prompting (FLP) method, which induces LLM to generate effective PDDL goals. Through comparative prompting experiments and systematic experiments, we demonstrate the effectiveness and robustness of LLM-PAS in handling anomalous conditions during task execution.
Simulating collective decision-making involves more than aggregating individual behaviors; it arises from dynamic interactions among individuals. While large language models (LLMs) show promise for social simulation, existing approaches often exhibit deviations from real-world data. To address this gap, we propose the Mean-Field LLM (MF-LLM) framework, which explicitly models the feedback loop between micro-level decisions and macro-level population. MF-LLM alternates between two models: a policy model that generates individual actions based on personal states and group-level information, and a mean field model that updates the population distribution from the latest individual decisions. Together, they produce rollouts that simulate the evolving trajectories of collective decision-making. To better match real-world data, we introduce IB-Tune, a fine-tuning method for LLMs grounded in the information bottleneck principle, which maximizes the relevance of population distributions to future actions while minimizing redundancy with historical data. We evaluate MF-LLM on a real-world social dataset, where it reduces KL divergence to human population distributions by 47 percent over non-mean-field baselines, and enables accurate trend forecasting and intervention planning. It generalizes across seven domains and four LLM backbones, providing a scalable foundation for high-fidelity social simulation.
A core challenge in AI-guided autonomy is enabling agents to navigate realistically and effectively in previously unseen environments based on natural language commands. We propose UAV-VLN, a novel end-to-end Vision-Language Navigation (VLN) framework for Unmanned Aerial Vehicles (UAVs) that seamlessly integrates Large Language Models (LLMs) with visual perception to facilitate human-interactive navigation. Our system interprets free-form natural language instructions, grounds them into visual observations, and plans feasible aerial trajectories in diverse environments. UAV-VLN leverages the common-sense reasoning capabilities of LLMs to parse high-level semantic goals, while a vision model detects and localizes semantically relevant objects in the environment. By fusing these modalities, the UAV can reason about spatial relationships, disambiguate references in human instructions, and plan context-aware behaviors with minimal task-specific supervision. To ensure robust and interpretable decision-making, the framework includes a cross-modal grounding mechanism that aligns linguistic intent with visual context. We evaluate UAV-VLN across diverse indoor and outdoor navigation scenarios, demonstrating its ability to generalize to novel instructions and environments with minimal task-specific training. Our results show significant improvements in instruction-following accuracy and trajectory efficiency, highlighting the potential of LLM-driven vision-language interfaces for safe, intuitive, and generalizable UAV autonomy.
As LLM-generated code grows in popularity, more evaluation is needed to assess the risks of using such tools, especially for safety-critical applications such as path planning. Existing coding benchmarks are insufficient as they do not reflect the context and complexity of safety-critical applications. To this end, we assessed six LLMs' abilities to generate the code for three different path-planning algorithms and tested them on three maps of various difficulties. Our results suggest that LLM-generated code presents serious hazards for path planning applications and should not be applied in safety-critical contexts without rigorous testing.
This study explores the integration of AI in transportation electrification planning in Austin, TX, focusing on the use of Geospatial AI (GeoAI), Generative AI (GenAI), and Large Language Models (LLMs). GeoAI enhances site selection, localized GenAI models support meta-level estimations, and LLMs enable scenario simulations. These AI applications require human oversight. GeoAI outputs must be evaluated with land use data, GenAI models are not always accurate, and LLMs are prone to hallucinations. To ensure accountable planning, human planners must work alongside AI agents. Establishing a community feedback loop is essential to audit automated decisions. Planners should place Community Experience (CX) at the center of Urban Planning AI.
LLM-integrated app systems extend the utility of Large Language Models (LLMs) with third-party apps that are invoked by a system LLM using interleaved planning and execution phases to answer user queries. These systems introduce new attack vectors where malicious apps can cause integrity violation of planning or execution, availability breakdown, or privacy compromise during execution. In this work, we identify new attacks impacting the integrity of planning, as well as the integrity and availability of execution in LLM-integrated apps, and demonstrate them against IsolateGPT, a recent solution designed to mitigate attacks from malicious apps. We propose Abstract-Concrete-Execute (ACE), a new secure architecture for LLM-integrated app systems that provides security guarantees for system planning and execution. Specifically, ACE decouples planning into two phases by first creating an abstract execution plan using only trusted information, and then mapping the abstract plan to a concrete plan using installed system apps. We verify that the plans generated by our system satisfy user-specified secure information flow constraints via static analysis on the structured plan output. During execution, ACE enforces data and capability barriers between apps, and ensures that the execution is conducted according to the trusted abstract plan. We show experimentally that our system is secure against attacks from the INJECAGENT benchmark, a standard benchmark for control flow integrity in the face of indirect prompt injection attacks, and our newly introduced attacks. Our architecture represents a significant advancement towards hardening LLM-based systems containing system facilities of varying levels of trustworthiness.
Given natural language test case description for an Android application, existing testing approaches require developers to manually write scripts using tools such as Appium and Espresso to execute the corresponding test case. This process is labor-intensive and demands significant effort to maintain as UI interfaces evolve throughout development. In this work, we introduce LELANTE, a novel framework that utilizes large language models (LLMs) to automate test case execution without requiring pre-written scripts. LELANTE interprets natural language test case descriptions, iteratively generate action plans, and perform the actions directly on the Android screen using its GUI. LELANTE employs a screen refinement process to enhance LLM interpretability, constructs a structured prompt for LLMs, and implements an action generation mechanism based on chain-of-thought reasoning of LLMs. To further reduce computational cost and enhance scalability, LELANTE utilizes model distillation using a foundational LLM. In experiments across 390 test cases spanning 10 popular Android applications, LELANTE achieved a 73% test execution success rate. Our results demonstrate that LLMs can effectively bridge the gap between natural language test case description and automated execution, making mobile testing more scalable and adaptable.
Graphical User Interface (GUI) agents, driven by Multi-modal Large Language Models (MLLMs), have emerged as a promising paradigm for enabling intelligent interaction with digital systems. This paper provides a structured summary of recent advances in GUI agents, focusing on architectures enhanced by Reinforcement Learning (RL). We first formalize GUI agent tasks as Markov Decision Processes and discuss typical execution environments and evaluation metrics. We then review the modular architecture of (M)LLM-based GUI agents, covering Perception, Planning, and Acting modules, and trace their evolution through representative works. Furthermore, we categorize GUI agent training methodologies into Prompt-based, Supervised Fine-Tuning (SFT)-based, and RL-based approaches, highlighting the progression from simple prompt engineering to dynamic policy learning via RL. Our summary illustrates how recent innovations in multimodal perception, decision reasoning, and adaptive action generation have significantly improved the generalization and robustness of GUI agents in complex real-world environments. We conclude by identifying key challenges and future directions for building more capable and reliable GUI agents.
We present CineVerse, a novel framework for the task of cinematic scene composition. Similar to traditional multi-shot generation, our task emphasizes the need for consistency and continuity across frames. However, our task also focuses on addressing challenges inherent to filmmaking, such as multiple characters, complex interactions, and visual cinematic effects. In order to learn to generate such content, we first create the CineVerse dataset. We use this dataset to train our proposed two-stage approach. First, we prompt a large language model (LLM) with task-specific instructions to take in a high-level scene description and generate a detailed plan for the overall setting and characters, as well as the individual shots. Then, we fine-tune a text-to-image generation model to synthesize high-quality visual keyframes. Experimental results demonstrate that CineVerse yields promising improvements in generating visually coherent and contextually rich movie scenes, paving the way for further exploration in cinematic video synthesis.
With the rapid rise of large language models (LLMs), phone automation has undergone transformative changes. This paper systematically reviews LLM-driven phone GUI agents, highlighting their evolution from script-based automation to intelligent, adaptive systems. We first contextualize key challenges, (i) limited generality, (ii) high maintenance overhead, and (iii) weak intent comprehension, and show how LLMs address these issues through advanced language understanding, multimodal perception, and robust decision-making. We then propose a taxonomy covering fundamental agent frameworks (single-agent, multi-agent, plan-then-act), modeling approaches (prompt engineering, training-based), and essential datasets and benchmarks. Furthermore, we detail task-specific architectures, supervised fine-tuning, and reinforcement learning strategies that bridge user intent and GUI operations. Finally, we discuss open challenges such as dataset diversity, on-device deployment efficiency, user-centric adaptation, and security concerns, offering forward-looking insights into this rapidly evolving field. By providing a structured overview and identifying pressing research gaps, this paper serves as a definitive reference for researchers and practitioners seeking to harness LLMs in designing scalable, user-friendly phone GUI agents.
A key factor for lunar mission planning is the ability to assess the local availability of raw materials. However, many potentially relevant measurements are scattered across a variety of scientific publications. In this paper we consider the viability of obtaining lunar composition data by leveraging LLMs to rapidly process a corpus of scientific publications. While leveraging LLMs to obtain knowledge from scientific documents is not new, this particular application presents interesting challenges due to the heterogeneity of lunar samples and the nuances involved in their characterization. Accuracy and uncertainty quantification are particularly crucial since many materials properties can be sensitive to small variations in composition. Our findings indicate that off-the-shelf LLMs are generally effective at extracting data from tables commonly found in these documents. However, there remains opportunity to further refine the data we extract in this initial approach; in particular, to capture fine-grained mineralogy information and to improve performance on more subtle/complex pieces of information.
In this paper we introduce ResearchCodeAgent, a novel multi-agent system leveraging large language models (LLMs) agents to automate the codification of research methodologies described in machine learning literature. The system bridges the gap between high-level research concepts and their practical implementation, allowing researchers auto-generating code of existing research papers for benchmarking or building on top-of existing methods specified in the literature with availability of partial or complete starter code. ResearchCodeAgent employs a flexible agent architecture with a comprehensive action suite, enabling context-aware interactions with the research environment. The system incorporates a dynamic planning mechanism, utilizing both short and long-term memory to adapt its approach iteratively. We evaluate ResearchCodeAgent on three distinct machine learning tasks with distinct task complexity and representing different parts of the ML pipeline: data augmentation, optimization, and data batching. Our results demonstrate the system's effectiveness and generalizability, with 46.9% of generated code being high-quality and error-free, and 25% showing performance improvements over baseline implementations. Empirical analysis shows an average reduction of 57.9% in coding time compared to manual implementation. We observe higher gains for more complex tasks. ResearchCodeAgent represents a significant step towards automating the research implementation process, potentially accelerating the pace of machine learning research.
In the field of large language model (LLM) post-training, the effectiveness of utilizing synthetic data generated by the LLM itself has been well-presented. However, a key question remains unaddressed: what essential information should such self-generated data encapsulate? Existing approaches only produce step-by-step problem solutions, and fail to capture the abstract meta-knowledge necessary for generalization across similar problems. Drawing insights from cognitive science, where humans employ high-level abstraction to simplify complex problems before delving into specifics, we introduce a novel self-training algorithm: LEarning to Plan before Answering (LEPA). LEPA trains the LLM to formulate anticipatory plans, which serve as abstract meta-knowledge for problem-solving, before engaging with the intricacies of problems. This approach not only outlines the solution generation path but also shields the LLM from the distraction of irrelevant details. During data generation, LEPA first crafts an anticipatory plan based on the problem, and then generates a solution that aligns with both the plan and the problem. LEPA refines the plan through self-reflection, aiming to acquire plans that are instrumental in yielding correct solutions. During model optimization, the LLM is trained to predict both the refined plans and the corresponding solutions. By efficiently extracting and utilizing the anticipatory plans, LEPA demonstrates remarkable superiority over conventional algorithms on various challenging natural language reasoning benchmarks.
Recent progress in large language models (LLMs) has shown promise in formal theorem proving, yet existing benchmarks remain limited to isolated, static proof tasks, failing to capture the iterative, engineering-intensive workflows of real-world formal mathematics libraries. Motivated by analogous advances in software engineering, we introduce the paradigm of Automated Proof Engineering (APE), which aims to automate proof engineering tasks such as feature addition, proof refactoring, and bug fixing using LLMs. To facilitate research in this direction, we present APE-Bench I, the first realistic benchmark built from real-world commit histories of Mathlib4, featuring diverse file-level tasks described in natural language and verified via a hybrid approach combining the Lean compiler and LLM-as-a-Judge. We further develop Eleanstic, a scalable parallel verification infrastructure optimized for proof checking across multiple versions of Mathlib. Empirical results on state-of-the-art LLMs demonstrate strong performance on localized edits but substantial degradation on handling complex proof engineering. This work lays the foundation for developing agentic workflows in proof engineering, with future benchmarks targeting multi-file coordination, project-scale verification, and autonomous agents capable of planning, editing, and repairing formal libraries.
Embodied systems, where generative autonomous agents engage with the physical world through integrated perception, cognition, action, and advanced reasoning powered by large language models (LLMs), hold immense potential for addressing complex, long-horizon, multi-objective tasks in real-world environments. However, deploying these systems remains challenging due to prolonged runtime latency, limited scalability, and heightened sensitivity, leading to significant system inefficiencies. In this paper, we aim to understand the workload characteristics of embodied agent systems and explore optimization solutions. We systematically categorize these systems into four paradigms and conduct benchmarking studies to evaluate their task performance and system efficiency across various modules, agent scales, and embodied tasks. Our benchmarking studies uncover critical challenges, such as prolonged planning and communication latency, redundant agent interactions, complex low-level control mechanisms, memory inconsistencies, exploding prompt lengths, sensitivity to self-correction and execution, sharp declines in success rates, and reduced collaboration efficiency as agent numbers increase. Leveraging these profiling insights, we suggest system optimization strategies to improve the performance, efficiency, and scalability of embodied agents across different paradigms. This paper presents the first system-level analysis of embodied AI agents, and explores opportunities for advancing future embodied system design.
Legal practitioners, particularly those early in their careers, face complex, high-stakes tasks that require adaptive, context-sensitive reasoning. While AI holds promise in supporting legal work, current datasets and models are narrowly focused on isolated subtasks and fail to capture the end-to-end decision-making required in real-world practice. To address this gap, we introduce LawFlow, a dataset of complete end-to-end legal workflows collected from trained law students, grounded in real-world business entity formation scenarios. Unlike prior datasets focused on input-output pairs or linear chains of thought, LawFlow captures dynamic, modular, and iterative reasoning processes that reflect the ambiguity, revision, and client-adaptive strategies of legal practice. Using LawFlow, we compare human and LLM-generated workflows, revealing systematic differences in structure, reasoning flexibility, and plan execution. Human workflows tend to be modular and adaptive, while LLM workflows are more sequential, exhaustive, and less sensitive to downstream implications. Our findings also suggest that legal professionals prefer AI to carry out supportive roles, such as brainstorming, identifying blind spots, and surfacing alternatives, rather than executing complex workflows end-to-end. Building on these findings, we propose a set of design suggestions, rooted in empirical observations, that align AI assistance with human goals of clarity, completeness, creativity, and efficiency, through hybrid planning, adaptive execution, and decision-point support. Our results highlight both the current limitations of LLMs in supporting complex legal workflows and opportunities for developing more collaborative, reasoning-aware legal AI systems. All data and code are available on our project page (https://minnesotanlp.github.io/LawFlow-website/).