Chain-of-Thought (CoT) prompting improves reasoning in large language models (LLMs), but its reliance on unstructured text limits interpretability and executability in embodied tasks. Prior work has explored structured CoTs using scene or logic graphs, yet these remain fundamentally limited: they model only low-order relations, lack constructs like inheritance or behavioral abstraction, and provide no standardized semantics for sequential or conditional planning. We propose UML-CoT, a structured reasoning and planning framework that leverages Unified Modeling Language (UML) to generate symbolic CoTs and executable action plans. UML class diagrams capture compositional object semantics, while activity diagrams model procedural control flow. Our three-stage training pipeline combines supervised fine-tuning with Group Relative Policy Optimization (GRPO), including reward learning from answer-only data. We evaluate UML-CoT on MRoom-30k, a new benchmark of cluttered room-cleaning scenarios. UML-CoT outperforms unstructured CoTs in interpretability, planning coherence, and execution success, highlighting UML as a more expressive and actionable structured reasoning formalism.
Recent reinforcement learning (RL) methods have substantially enhanced the planning capabilities of Large Language Models (LLMs), yet the theoretical basis for their effectiveness remains elusive. In this work, we investigate RL's benefits and limitations through a tractable graph-based abstraction, focusing on policy gradient (PG) and Q-learning methods. Our theoretical analyses reveal that supervised fine-tuning (SFT) may introduce co-occurrence-based spurious solutions, whereas RL achieves correct planning primarily through exploration, underscoring exploration's role in enabling better generalization. However, we also show that PG suffers from diversity collapse, where output diversity decreases during training and persists even after perfect accuracy is attained. By contrast, Q-learning provides two key advantages: off-policy learning and diversity preservation at convergence. We further demonstrate that careful reward design is necessary to prevent reward hacking in Q-learning. Finally, applying our framework to the real-world planning benchmark Blocksworld, we confirm that these behaviors manifest in practice.
LLMs trained for logical reasoning excel at step-by-step deduction to reach verifiable answers. However, this paradigm is ill-suited for navigating social situations, which induce an interpretive process of analyzing ambiguous cues that rarely yield a definitive outcome. To bridge this gap, we introduce Cognitive Reasoning, a paradigm modeled on human social cognition. It formulates the interpretive process into a structured cognitive flow of interconnected cognitive units (e.g., observation or attribution), which combine adaptively to enable effective social thinking and responses. We then propose CogFlow, a complete framework that instills this capability in LLMs. CogFlow first curates a dataset of cognitive flows by simulating the associative and progressive nature of human thought via tree-structured planning. After instilling the basic cognitive reasoning capability via supervised fine-tuning, CogFlow adopts reinforcement learning to enable the model to improve itself via trial and error, guided by a multi-objective reward that optimizes both cognitive flow and response quality. Extensive experiments show that CogFlow effectively enhances the social cognitive capabilities of LLMs, and even humans, leading to more effective social decision-making.
Automatically reproducing Android app crashes from textual bug reports is challenging, particularly when the reports are incomplete and the modern UI exhibits high combinatorial complexity. Existing approaches based on reinforcement learning or large language models (LLMs) exhibit limitations in such scenarios. They struggle to infer unobserved steps and reconstruct the underlying user action sequences to navigate the vast UI interaction space, primarily due to limited goal-directed reasoning and planning. We present TreeMind, a novel technique that integrates LLMs with a customized Monte Carlo Tree Search (MCTS) algorithm to achieve strategic UI exploration in bug reproduction. To the best of our knowledge, this is the first work to combine external decision-making with LLM semantic reasoning for reliable bug reproduction. We formulate the reproduction task as a target-driven search problem, leveraging MCTS as the core planning mechanism to iteratively refine action sequences. To enhance MCTS with semantic reasoning, we introduce two LLM-guided agents with distinct roles: Expander generates top-k promising actions based on the current UI state and exploration history, while Simulator estimates the likelihood that each action leads toward successful reproduction. By incorporating multi-modal UI inputs and advanced prompting techniques, TreeMind conducts feedback-aware navigation that identifies missing but essential user actions and incrementally reconstructs the reproduction paths. We evaluate TreeMind on a dataset of 93 real-world Android bug reports from three widely-used benchmarks. Experimental results show that it significantly outperforms four state-of-the-art baselines in reproduction success rate. A real-world case study indicates that integrating LLM reasoning with MCTS-based planning is a compelling direction for automated bug reproduction.
Inspired by the dual-process theory of human cognition from \textit{Thinking, Fast and Slow}, we introduce \textbf{PRIME} (Planning and Retrieval-Integrated Memory for Enhanced Reasoning), a multi-agent reasoning framework that dynamically integrates \textbf{System 1} (fast, intuitive thinking) and \textbf{System 2} (slow, deliberate thinking). PRIME first employs a Quick Thinking Agent (System 1) to generate a rapid answer; if uncertainty is detected, it then triggers a structured System 2 reasoning pipeline composed of specialized agents for \textit{planning}, \textit{hypothesis generation}, \textit{retrieval}, \textit{information integration}, and \textit{decision-making}. This multi-agent design faithfully mimics human cognitive processes and enhances both efficiency and accuracy. Experimental results with LLaMA 3 models demonstrate that PRIME enables open-source LLMs to perform competitively with state-of-the-art closed-source models like GPT-4 and GPT-4o on benchmarks requiring multi-hop and knowledge-grounded reasoning. This research establishes PRIME as a scalable solution for improving LLMs in domains requiring complex, knowledge-intensive reasoning.
GUI task automation streamlines repetitive tasks, but existing LLM or VLM-based planner-executor agents suffer from brittle generalization, high latency, and limited long-horizon coherence. Their reliance on single-shot reasoning or static plans makes them fragile under UI changes or complex tasks. Log2Plan addresses these limitations by combining a structured two-level planning framework with a task mining approach over user behavior logs, enabling robust and adaptable GUI automation. Log2Plan constructs high-level plans by mapping user commands to a structured task dictionary, enabling consistent and generalizable automation. To support personalization and reuse, it employs a task mining approach from user behavior logs that identifies user-specific patterns. These high-level plans are then grounded into low-level action sequences by interpreting real-time GUI context, ensuring robust execution across varying interfaces. We evaluated Log2Plan on 200 real-world tasks, demonstrating significant improvements in task success rate and execution time. Notably, it maintains over 60.0% success rate even on long-horizon task sequences, highlighting its robustness in complex, multi-step workflows.
Chain-of-Thought (CoT) prompting helps Large Language Models (LLMs) tackle complex reasoning by eliciting explicit step-by-step rationales. However, CoT's verbosity increases latency and memory usage and may propagate early errors across long chains. We propose the Reasoning Capsule (R-Capsule), a framework that aims to combine the efficiency of latent reasoning with the transparency of explicit CoT. The core idea is to compress the high-level plan into a small set of learned latent tokens (a Reasoning Capsule) while keeping execution steps lightweight or explicit. This hybrid approach is inspired by the Information Bottleneck (IB) principle, where we encourage the capsule to be approximately minimal yet sufficient for the task. Minimality is encouraged via a low-capacity bottleneck, which helps improve efficiency. Sufficiency is encouraged via a dual objective: a primary task loss for answer accuracy and an auxiliary plan-reconstruction loss that encourages the capsule to faithfully represent the original textual plan. The reconstruction objective helps ground the latent space, thereby improving interpretability and reducing the use of uninformative shortcuts. Our framework strikes a balance between efficiency, accuracy, and interpretability, thereby reducing the visible token footprint of reasoning while maintaining or improving accuracy on complex benchmarks. Our codes are available at: https://anonymous.4open.science/r/Reasoning-Capsule-7BE0
Multi-Agent Path Finding (MAPF) poses a significant and challenging problem critical for applications in robotics and logistics, particularly due to its combinatorial complexity and the partial observability inherent in realistic environments. Decentralized reinforcement learning methods commonly encounter two substantial difficulties: first, they often yield self-centered behaviors among agents, resulting in frequent collisions, and second, their reliance on complex communication modules leads to prolonged training times, sometimes spanning weeks. To address these challenges, we propose an efficient decentralized planning framework based on the Decision Transformer (DT), uniquely leveraging offline reinforcement learning to substantially reduce training durations from weeks to mere hours. Crucially, our approach effectively handles long-horizon credit assignment and significantly improves performance in scenarios with sparse and delayed rewards. Furthermore, to overcome adaptability limitations inherent in standard RL methods under dynamic environmental changes, we integrate a large language model (GPT-4o) to dynamically guide agent policies. Extensive experiments in both static and dynamically changing environments demonstrate that our DT-based approach, augmented briefly by GPT-4o, significantly enhances adaptability and performance.
Existing Large Language Model (LLM) based autoregressive (AR) text-to-speech (TTS) systems, while achieving state-of-the-art quality, still face critical challenges. The foundation of this LLM-based paradigm is the discretization of the continuous speech waveform into a sequence of discrete tokens by neural audio codec. However, single codebook modeling is well suited to text LLMs, but suffers from significant information loss; hierarchical acoustic tokens, typically generated via Residual Vector Quantization (RVQ), often lack explicit semantic structure, placing a heavy learning burden on the model. Furthermore, the autoregressive process is inherently susceptible to error accumulation, which can degrade generation stability. To address these limitations, we propose CaT-TTS, a novel framework for robust and semantically-grounded zero-shot synthesis. First, we introduce S3Codec, a split RVQ codec that injects explicit linguistic features into its primary codebook via semantic distillation from a state-of-the-art ASR model, providing a structured representation that simplifies the learning task. Second, we propose an ``Understand-then-Generate'' dual-Transformer architecture that decouples comprehension from rendering. An initial ``Understanding'' Transformer models the cross-modal relationship between text and the audio's semantic tokens to form a high-level utterance plan. A subsequent ``Generation'' Transformer then executes this plan, autoregressively synthesizing hierarchical acoustic tokens. Finally, to enhance generation stability, we introduce Masked Audio Parallel Inference (MAPI), a nearly parameter-free inference strategy that dynamically guides the decoding process to mitigate local errors.
Real-world decision-making tasks typically occur in complex and open environments, posing significant challenges to reinforcement learning (RL) agents' exploration efficiency and long-horizon planning capabilities. A promising approach is LLM-enhanced RL, which leverages the rich prior knowledge and strong planning capabilities of LLMs to guide RL agents in efficient exploration. However, existing methods mostly rely on frequent and costly LLM invocations and suffer from limited performance due to the semantic mismatch. In this paper, we introduce a Structured Goal-guided Reinforcement Learning (SGRL) method that integrates a structured goal planner and a goal-conditioned action pruner to guide RL agents toward efficient exploration. Specifically, the structured goal planner utilizes LLMs to generate a reusable, structured function for goal generation, in which goals are prioritized. Furthermore, by utilizing LLMs to determine goals' priority weights, it dynamically generates forward-looking goals to guide the agent's policy toward more promising decision-making trajectories. The goal-conditioned action pruner employs an action masking mechanism that filters out actions misaligned with the current goal, thereby constraining the RL agent to select goal-consistent policies. We evaluate the proposed method on Crafter and Craftax-Classic, and experimental results demonstrate that SGRL achieves superior performance compared to existing state-of-the-art methods.
Effective real-world multi-agent collaboration requires not only accurate planning but also the ability to reason about collaborators' intents -- a crucial capability for avoiding miscoordination and redundant communication under partial observable environments. Due to their strong planning and reasoning capabilities, large language models (LLMs) have emerged as promising autonomous agents for collaborative task solving. However, existing collaboration frameworks for LLMs overlook their reasoning potential for dynamic intent inference, and thus produce inconsistent plans and redundant communication, reducing collaboration efficiency. To bridge this gap, we propose CoBel-World, a novel framework that equips LLM agents with a collaborative belief world -- an internal representation jointly modeling the physical environment and collaborators' mental states. CoBel-World enables agents to parse open-world task knowledge into structured beliefs via a symbolic belief language, and perform zero-shot Bayesian-style belief updates through LLM reasoning. This allows agents to proactively detect potential miscoordination (e.g., conflicting plans) and communicate adaptively. Evaluated on challenging embodied benchmarks (i.e., TDW-MAT and C-WAH), CoBel-World significantly reduces communication costs by 22-60% and improves task completion efficiency by 4-28% compared to the strongest baseline. Our results show that explicit, intent-aware belief modeling is essential for efficient and human-like collaboration in LLM-based multi-agent systems.
Successfully solving long-horizon manipulation tasks remains a fundamental challenge. These tasks involve extended action sequences and complex object interactions, presenting a critical gap between high-level symbolic planning and low-level continuous control. To bridge this gap, two essential capabilities are required: robust long-horizon task planning and effective goal-conditioned manipulation. Existing task planning methods, including traditional and LLM-based approaches, often exhibit limited generalization or sparse semantic reasoning. Meanwhile, image-conditioned control methods struggle to adapt to unseen tasks. To tackle these problems, we propose SAGE, a novel framework for Scene Graph-Aware Guidance and Execution in Long-Horizon Manipulation Tasks. SAGE utilizes semantic scene graphs as a structural representation for scene states. A structural scene graph enables bridging task-level semantic reasoning and pixel-level visuo-motor control. This also facilitates the controllable synthesis of accurate, novel sub-goal images. SAGE consists of two key components: (1) a scene graph-based task planner that uses VLMs and LLMs to parse the environment and reason about physically-grounded scene state transition sequences, and (2) a decoupled structural image editing pipeline that controllably converts each target sub-goal graph into a corresponding image through image inpainting and composition. Extensive experiments have demonstrated that SAGE achieves state-of-the-art performance on distinct long-horizon tasks.
There is growing interest in using Large Language Models as agents (LLM agents) for social simulations to inform policy, yet real-world adoption remains limited. This paper addresses the question: How can LLM agent simulations be made genuinely useful for policy? We report on a year-long iterative design engagement with a university emergency preparedness team. Across multiple design iterations, we iteratively developed a system of 13,000 LLM agents that simulate crowd movement and communication during a large-scale gathering under various emergency scenarios. These simulations informed actual policy implementation, shaping volunteer training, evacuation protocols, and infrastructure planning. Analyzing this process, we identify three design implications: start with verifiable scenarios and build trust gradually, use preliminary simulations to elicit tacit knowledge, and treat simulation and policy development as evolving together. These implications highlight actionable pathways to making LLM agent simulations that are genuinely useful for policy.
Travel planning (TP) agent has recently worked as an emerging building block to interact with external tools and resources for travel itinerary generation, ensuring enjoyable user experience. Despite its benefits, existing studies rely on hand craft prompt and fixed agent workflow, hindering more flexible and autonomous TP agent. This paper proposes DeepTravel, an end to end agentic reinforcement learning framework for building autonomous travel planning agent, capable of autonomously planning, executing tools, and reflecting on tool responses to explore, verify, and refine intermediate actions in multi step reasoning. To achieve this, we first construct a robust sandbox environment by caching transportation, accommodation and POI data, facilitating TP agent training without being constrained by real world APIs limitations (e.g., inconsistent outputs). Moreover, we develop a hierarchical reward modeling system, where a trajectory level verifier first checks spatiotemporal feasibility and filters unsatisfied travel itinerary, and then the turn level verifier further validate itinerary detail consistency with tool responses, enabling efficient and precise reward service. Finally, we propose the reply augmented reinforcement learning method that enables TP agent to periodically replay from a failures experience buffer, emerging notable agentic capacity. We deploy trained TP agent on DiDi Enterprise Solutions App and conduct comprehensive online and offline evaluations, demonstrating that DeepTravel enables small size LLMs (e.g., Qwen3 32B) to significantly outperform existing frontier LLMs such as OpenAI o1, o3 and DeepSeek R1 in travel planning tasks.
Data science, which transforms raw data into actionable insights, is critical for data-driven decision-making. However, these tasks are often complex, involving steps for exploring multiple data sources and synthesizing findings to deliver insightful answers. While large language models (LLMs) show significant promise in automating this process, they often struggle with heterogeneous data formats and generate sub-optimal analysis plans, as verifying plan sufficiency is inherently difficult without ground-truth labels for such open-ended tasks. To overcome these limitations, we introduce DS-STAR, a novel data science agent. Specifically, DS-STAR makes three key contributions: (1) a data file analysis module that automatically explores and extracts context from diverse data formats, including unstructured types; (2) a verification step where an LLM-based judge evaluates the sufficiency of the analysis plan at each stage; and (3) a sequential planning mechanism that starts with a simple, executable plan and iteratively refines it based on the DS-STAR's feedback until its sufficiency is verified. This iterative refinement allows DS-STAR to reliably navigate complex analyses involving diverse data sources. Our experiments show that DS-STAR achieves state-of-the-art performance across three challenging benchmarks: DABStep, KramaBench, and DA-Code. Moreover, DS-STAR particularly outperforms baselines on hard tasks that require processing multiple data files with heterogeneous formats.
Excel is one of the most widely used productivity tools across domains, offering rich functionality but also overwhelming users with its complexity. This creates a persistent demand for tutorials to support effective usage. However, existing tutorials are manually authored by experts, require frequent updates after each software release, and incur substantial labor costs. Prior work has not achieved fully automated tutorial generation, since existing methods still depend on handcrafted operation sequences or example materials. In this paper, we present the first framework for automatically generating Excel tutorials directly from natural language task descriptions. Our framework first instantiates the task. Then a central component of this framework, Execution Agent, plans and executes the solution in Excel, and collects the intermediate artifacts required for tutorial construction. These artifacts are then transformed into both structured Excel documents and video demonstrations. To build a comprehensive tutorial corpus, we collected 1,559 task descriptions from real-world scenarios. In addition, we designed a systematic evaluation framework that integrates assessments from both large language models (LLMs) and human reviewers. Experimental results show that our framework improves task execution success rates by 8.5% over state-of-the-art baselines. Moreover, the generated tutorials demonstrate superior readability and instructional effectiveness, often approaching or surpassing expert-authored materials. Importantly, the automated pipeline eliminates manual labor and reduces time costs to 1/20 of expert authoring, making scalable and high-quality tutorial generation practical for the first time.
Autonomous agents have recently achieved remarkable progress across diverse domains, yet most evaluations focus on short-horizon, fully observable tasks. In contrast, many critical real-world tasks, such as large-scale software development, commercial investment, and scientific discovery, unfold in long-horizon and partially observable scenarios where success hinges on sustained reasoning, planning, memory management, and tool use. Existing benchmarks rarely capture these long-horizon challenges, leaving a gap in systematic evaluation. To bridge this gap, we introduce \textbf{UltraHorizon} a novel benchmark that measures the foundational capabilities essential for complex real-world challenges. We use exploration as a unifying task across three distinct environments to validate these core competencies. Agents are designed in long-horizon discovery tasks where they must iteratively uncover hidden rules through sustained reasoning, planning, memory and tools management, and interaction with environments. Under the heaviest scale setting, trajectories average \textbf{200k+} tokens and \textbf{400+} tool calls, whereas in standard configurations they still exceed \textbf{35k} tokens and involve more than \textbf{60} tool calls on average. Our extensive experiments reveal that LLM-agents consistently underperform in these settings, whereas human participants achieve higher scores, underscoring a persistent gap in agents' long-horizon abilities. We also observe that simple scaling fails in our task. To better illustrate the failure of agents, we conduct an in-depth analysis of collected trajectories. We identify eight types of errors and attribute them to two primary causes: in-context locking and functional fundamental capability gaps. \href{https://github.com/StarDewXXX/UltraHorizon}{Our code will be available here.}
We introduce QueryGym, an interactive environment for building, testing, and evaluating LLM-based query planning agents. Existing frameworks often tie agents to specific query language dialects or obscure their reasoning; QueryGym instead requires agents to construct explicit sequences of relational algebra operations, ensuring engine-agnostic evaluation and transparent step-by-step planning. The environment is implemented as a Gymnasium interface that supplies observations -- including schema details, intermediate results, and execution feedback -- and receives actions that represent database exploration (e.g., previewing tables, sampling column values, retrieving unique values) as well as relational algebra operations (e.g., filter, project, join). We detail the motivation and the design of the environment. In the demo, we showcase the utility of the environment by contrasting it with contemporary LLMs that query databases. QueryGym serves as a practical testbed for research in error remediation, transparency, and reinforcement learning for query generation. For the associated demo, see https://ibm.biz/QueryGym.
Multimodal Procedural Planning (MPP) aims to generate step-by-step instructions that combine text and images, with the central challenge of preserving object-state consistency across modalities while producing informative plans. Existing approaches often leverage large language models (LLMs) to refine textual steps; however, visual object-state alignment and systematic evaluation are largely underexplored. We present MMPlanner, a zero-shot MPP framework that introduces Object State Reasoning Chain-of-Thought (OSR-CoT) prompting to explicitly model object-state transitions and generate accurate multimodal plans. To assess plan quality, we design LLM-as-a-judge protocols for planning accuracy and cross-modal alignment, and further propose a visual step-reordering task to measure temporal coherence. Experiments on RECIPEPLAN and WIKIPLAN show that MMPlanner achieves state-of-the-art performance, improving textual planning by +6.8%, cross-modal alignment by +11.9%, and visual step ordering by +26.7%
The advancement of vision language models (VLMs) has empowered embodied agents to accomplish simple multimodal planning tasks, but not long-horizon ones requiring long sequences of actions. In text-only simulations, long-horizon planning has seen significant improvement brought by repositioning the role of LLMs. Instead of directly generating action sequences, LLMs translate the planning domain and problem into a formal planning language like the Planning Domain Definition Language (PDDL), which can call a formal solver to derive the plan in a verifiable manner. In multimodal environments, research on VLM-as-formalizer remains scarce, usually involving gross simplifications such as predefined object vocabulary or overly similar few-shot examples. In this work, we present a suite of five VLM-as-formalizer pipelines that tackle one-shot, open-vocabulary, and multimodal PDDL formalization. We evaluate those on an existing benchmark while presenting another two that for the first time account for planning with authentic, multi-view, and low-quality images. We conclude that VLM-as-formalizer greatly outperforms end-to-end plan generation. We reveal the bottleneck to be vision rather than language, as VLMs often fail to capture an exhaustive set of necessary object relations. While generating intermediate, textual representations such as captions or scene graphs partially compensate for the performance, their inconsistent gain leaves headroom for future research directions on multimodal planning formalization.
Large Language Models (LLMs) have recently shown strong potential in robotic task planning, particularly through automatic planning domain generation that integrates symbolic search. Prior approaches, however, have largely treated these domains as search utilities, with limited attention to their potential as scalable sources of reasoning data. At the same time, progress in reasoning LLMs has been driven by chain-of-thought (CoT) supervision, whose application in robotics remains dependent on costly, human-curated datasets. We propose Plan2Evolve, an LLM self-evolving framework in which the base model generates planning domains that serve as engines for producing symbolic problem-plan pairs as reasoning traces. These pairs are then transformed into extended CoT trajectories by the same model through natural-language explanations, thereby explicitly aligning symbolic planning structures with natural language reasoning. The resulting data extend beyond the model's intrinsic planning capacity, enabling model fine-tuning that yields a planning-enhanced LLM with improved planning success, stronger cross-task generalization, and reduced inference costs.
Retrosynthesis planning enables the discovery of viable synthetic routes for target molecules, playing a crucial role in domains like drug discovery and materials design. Multi-step retrosynthetic planning remains computationally challenging due to exponential search spaces and inference costs. While Large Language Models (LLMs) demonstrate chemical reasoning capabilities, their application to synthesis planning faces constraints on efficiency and cost. To address these challenges, we introduce AOT*, a framework that transforms retrosynthetic planning by integrating LLM-generated chemical synthesis pathways with systematic AND-OR tree search. To this end, AOT* atomically maps the generated complete synthesis routes onto AND-OR tree components, with a mathematically sound design of reward assignment strategy and retrieval-based context engineering, thus enabling LLMs to efficiently navigate in the chemical space. Experimental evaluation on multiple synthesis benchmarks demonstrates that AOT* achieves SOTA performance with significantly improved search efficiency. AOT* exhibits competitive solve rates using 3-5$\times$ fewer iterations than existing LLM-based approaches, with the efficiency advantage becoming more pronounced on complex molecular targets.
The goal of personalized medicine is to discover a treatment regimen that optimizes a patient's clinical outcome based on their personal genetic and environmental factors. However, candidate treatments cannot be arbitrarily administered to the patient to assess their efficacy; we often instead have access to an in silico surrogate model that approximates the true fitness of a proposed treatment. Unfortunately, such surrogate models have been shown to fail to generalize to previously unseen patient-treatment combinations. We hypothesize that domain-specific prior knowledge - such as medical textbooks and biomedical knowledge graphs - can provide a meaningful alternative signal of the fitness of proposed treatments. To this end, we introduce LLM-based Entropy-guided Optimization with kNowledgeable priors (LEON), a mathematically principled approach to leverage large language models (LLMs) as black-box optimizers without any task-specific fine-tuning, taking advantage of their ability to contextualize unstructured domain knowledge to propose personalized treatment plans in natural language. In practice, we implement LEON via 'optimization by prompting,' which uses LLMs as stochastic engines for proposing treatment designs. Experiments on real-world optimization tasks show LEON outperforms both traditional and LLM-based methods in proposing individualized treatments for patients.
Conventional SLAM pipelines for legged robot navigation are fragile under rapid motion, calibration demands, and sensor drift, while offering limited semantic reasoning for task-driven exploration. To deal with these issues, we propose a vision-only, SLAM-free navigation framework that replaces dense geometry with semantic reasoning and lightweight topological representations. A hierarchical vision-language perception module fuses scene-level context with object-level cues for robust semantic inference. And a semantic-probabilistic topological map supports coarse-to-fine planning: LLM-based global reasoning for subgoal selection and vision-based local planning for obstacle avoidance. Integrated with reinforcement-learning locomotion controllers, the framework is deployable across diverse legged robot platforms. Experiments in simulation and real-world settings demonstrate consistent improvements in semantic accuracy, planning quality, and navigation success, while ablation studies further showcase the necessity of both hierarchical perception and fine local planning. This work introduces a new paradigm for SLAM-free, vision-language-driven navigation, shifting robotic exploration from geometry-centric mapping to semantics-driven decision making.
Integrating natural language (NL) prompts into robotic mission planning has attracted significant interest in recent years. In the construction domain, Building Information Models (BIM) encapsulate rich NL descriptions of the environment. We present a novel framework that fuses NL directives with BIM-derived semantic maps via a Beta-Bernoulli Bayesian fusion by interpreting the LLM as a sensor: each obstacle's design-time repulsive coefficient is treated as a Beta(alpha, beta) random variable and LLM-returned danger scores are incorporated as pseudo-counts to update alpha and beta. The resulting posterior mean yields a continuous, context-aware repulsive gain that augments a Euclidean-distance-based potential field for cost heuristics. By adjusting gains based on sentiment and context inferred from user prompts, our method guides robots along safer, more context-aware paths. This provides a numerically stable method that can chain multiple natural commands and prompts from construction workers and foreman to enable planning while giving flexibility to be integrated in any learned or classical AI framework. Simulation results demonstrate that this Beta-Bernoulli fusion yields both qualitative and quantitative improvements in path robustness and validity.
Purpose: To develop a retrieval-augmented generation (RAG) system powered by LLaMA-4 109B for automated, protocol-aware, and interpretable evaluation of radiotherapy treatment plans. Methods and Materials: We curated a multi-protocol dataset of 614 radiotherapy plans across four disease sites and constructed a knowledge base containing normalized dose metrics and protocol-defined constraints. The RAG system integrates three core modules: a retrieval engine optimized across five SentenceTransformer backbones, a percentile prediction component based on cohort similarity, and a clinical constraint checker. These tools are directed by a large language model (LLM) using a multi-step prompt-driven reasoning pipeline to produce concise, grounded evaluations. Results: Retrieval hyperparameters were optimized using Gaussian Process on a scalarized loss function combining root mean squared error (RMSE), mean absolute error (MAE), and clinically motivated accuracy thresholds. The best configuration, based on all-MiniLM-L6-v2, achieved perfect nearest-neighbor accuracy within a 5-percentile-point margin and a sub-2pt MAE. When tested end-to-end, the RAG system achieved 100% agreement with the computed values by standalone retrieval and constraint-checking modules on both percentile estimates and constraint identification, confirming reliable execution of all retrieval, prediction and checking steps. Conclusion: Our findings highlight the feasibility of combining structured population-based scoring with modular tool-augmented reasoning for transparent, scalable plan evaluation in radiation therapy. The system offers traceable outputs, minimizes hallucination, and demonstrates robustness across protocols. Future directions include clinician-led validation, and improved domain-adapted retrieval models to enhance real-world integration.
Large Language Models (LLMs) have demonstrated remarkable capabilities in knowledge acquisition, reasoning, and tool use, making them promising candidates for autonomous agent applications. However, training LLM agents for complex multi-turn task planning faces significant challenges, including sparse episode-wise rewards, credit assignment across long horizons, and the computational overhead of reinforcement learning in multi-turn interaction settings. To this end, this paper introduces a novel approach that transforms multi-turn task planning into single-turn task reasoning problems, enabling efficient policy optimization through Group Relative Policy Optimization (GRPO) with dense and verifiable reward from expert trajectories. Our theoretical analysis shows that GRPO improvement on single-turn task reasoning results in higher multi-turn success probability under the minimal turns, as well as the generalization to subtasks with shorter horizons. Experimental evaluation on the complex task planning benchmark demonstrates that our 1.5B parameter model trained with single-turn GRPO achieves superior performance compared to larger baseline models up to 14B parameters, with success rates of 70% for long-horizon planning tasks with over 30 steps. We also theoretically and empirically validate the strong cross-task generalizability that the models trained on complex tasks can lead to the successful completion of all simpler subtasks.
Despite the rapid advancements in LLM agents, they still face the challenge of generating meaningful reflections due to inadequate error analysis and a reliance on rare successful trajectories, especially in complex tasks. In this work, we propose SAMULE, a new framework for self-learning agents powered by a retrospective language model that is trained based on Multi-Level Reflection Synthesis. It first synthesizes high-quality reflections across three complementary levels: Single-Trajectory Learning (micro-level) for detailed error correction; Intra-Task Learning (meso-level) to build error taxonomies across multiple trials of the same task, and Inter-Task Learning (macro-level) to extract transferable insights based on same typed errors from diverse task failures. Then we fine-tune a language model serving as the retrospective model to generate reflections during inference. We further extend our framework to interactive settings through a foresight-based reflection mechanism, enabling agents to proactively reflect and adapt during user interactions by comparing predicted and actual responses. Extensive experiments on three challenging benchmarks - TravelPlanner, NATURAL PLAN, and Tau-bench - demonstrate that our approach significantly outperforms reflection-based baselines. Our results highlight the critical role of well-designed reflection synthesis and failure-centric learning in building self-improving LLM agents.
As AI inference scales to billions of queries and emerging reasoning and agentic workflows increase token demand, reliable estimates of per-query energy use are increasingly important for capacity planning, emissions accounting, and efficiency prioritization. Many public estimates are inconsistent and overstate energy use, because they extrapolate from limited benchmarks and fail to reflect efficiency gains achievable at scale. In this perspective, we introduce a bottom-up methodology to estimate the per-query energy of large-scale LLM systems based on token throughput. For models running on an H100 node under realistic workloads, GPU utilization and PUE constraints, we estimate a median energy per query of 0.34 Wh (IQR: 0.18-0.67) for frontier-scale models (>200 billion parameters). These results are consistent with measurements using production-scale configurations and show that non-production estimates and assumptions can overstate energy use by 4-20x. Extending to test-time scaling scenarios with 15x more tokens per typical query, the median energy rises 13x to 4.32 Wh, indicating that targeting efficiency in this regime will deliver the largest fleet-wide savings. We quantify achievable efficiency gains at the model, serving platform, and hardware levels, finding individual median reductions of 1.5-3.5x in energy per query, while combined advances can plausibly deliver 8-20x reductions. To illustrate the system-level impact, we estimate the baseline daily energy use of a deployment serving 1 billion queries to be 0.8 GWh/day. If 10% are long queries, demand could grow to 1.8 GWh/day. With targeted efficiency interventions, it falls to 0.9 GWh/day, similar to the energy footprint of web search at that scale. This echoes how data centers historically tempered energy growth through efficiency gains during the internet and cloud build-up.
We present LLM Trainer, a fully automated pipeline that leverages the world knowledge of Large Language Models (LLMs) to transform a small number of human demonstrations (as few as one) into a large robot dataset for imitation learning. Our approach decomposes demonstration generation into two steps: (1) offline demonstration annotation that extracts keyframes, salient objects, and pose-object relations; and (2) online keypose retargeting that adapts those keyframes to a new scene, given an initial observation. Using these modified keypoints, our system warps the original demonstration to generate a new trajectory, which is then executed, and the resulting demo, if successful, is saved. Because the annotation is reusable across scenes, we use Thompson sampling to optimize the annotation, significantly improving generation success rate. We evaluate our method on a range of tasks, and find that our data annotation method consistently outperforms expert-engineered baselines. We further show an ensemble policy that combines the optimized LLM feed-forward plan with a learned feedback imitation learning controller. Finally, we demonstrate hardware feasibility on a Franka Emika Panda robot. For additional materials and demonstration videos, please see the project website: https://sites.google.com/andrew.cmu.edu/llm-trainer