Large language models (LLMs) are moving beyond static uses and are now powering agents that learn continually during their interaction with external environments. For example, agents can learn reusable skills while navigating web pages or toggling new tools. However, existing methods for skill learning often create skills that are over-specialized to a single website and fail to generalize. We introduce PolySkill, a new framework that enables agents to learn generalizable and compositional skills. The core idea, inspired by polymorphism in software engineering, is to decouple a skill's abstract goal (what it accomplishes) and its concrete implementation (how it is executed). Experiments show that our method (1) improves skill reuse by 1.7x on seen websites and (2) boosts success rates by up to 9.4% on Mind2Web and 13.9% on unseen websites, while reducing steps by over 20%. (3) In self-exploration settings without specified tasks, our framework improves the quality of proposed tasks and enables agents to learn generalizable skills that work across different sites. By enabling the agent to identify and refine its own goals, the PolySkill enhances the agent's ability to learn a better curriculum, leading to the acquisition of more generalizable skills compared to baseline methods. This work provides a practical path toward building agents capable of continual learning in adaptive environments. Our findings show that separating a skill's goal from its execution is a crucial step toward developing autonomous agents that can learn and generalize across the open web continuously.
Tool-augmented large language models (LLMs) are emerging as deep research agents, systems that decompose complex queries, retrieve external evidence, and synthesize grounded responses. Yet current agents remain limited by shallow retrieval, weak alignment metrics, and brittle tool-use behavior. We introduce PokeeResearch-7B, a 7B-parameter deep research agent built under a unified reinforcement learning framework for robustness, alignment, and scalability. PokeeResearch-7B is trained by an annotation-free Reinforcement Learning from AI Feedback (RLAIF) framework to optimize policies using LLM-based reward signals that capture factual accuracy, citation faithfulness, and instruction adherence. A chain-of-thought-driven multi-call reasoning scaffold further enhances robustness through self-verification and adaptive recovery from tool failures. Among 10 popular deep research benchmarks, PokeeResearch-7B achieves state-of-the-art performance among 7B-scale deep research agents. This highlights that careful reinforcement learning and reasoning design can produce efficient, resilient, and research-grade AI agents. The model and inference code is open-sourced under MIT license at https://github.com/Pokee-AI/PokeeResearchOSS.
Academic project websites can more effectively disseminate research when they clearly present core content and enable intuitive navigation and interaction. However, current approaches such as direct Large Language Model (LLM) generation, templates, or direct HTML conversion struggle to produce layout-aware, interactive sites, and a comprehensive evaluation suite for this task has been lacking. In this paper, we introduce Paper2Web, a benchmark dataset and multi-dimensional evaluation framework for assessing academic webpage generation. It incorporates rule-based metrics like Connectivity, Completeness and human-verified LLM-as-a-Judge (covering interactivity, aesthetics, and informativeness), and PaperQuiz, which measures paper-level knowledge retention. We further present PWAgent, an autonomous pipeline that converts scientific papers into interactive and multimedia-rich academic homepages. The agent iteratively refines both content and layout through MCP tools that enhance emphasis, balance, and presentation quality. Our experiments show that PWAgent consistently outperforms end-to-end baselines like template-based webpages and arXiv/alphaXiv versions by a large margin while maintaining low cost, achieving the Pareto-front in academic webpage generation.
So-called `wicked problems', those involving complex multi-dimensional settings, non-verifiable outcomes, heterogeneous impacts and a lack of single objectively correct answers, have plagued humans throughout history. Modern examples include decisions over justice frameworks, solving environmental pollution, planning for pandemic resilience and food security. The use of state-of-the-art artificial intelligence systems (notably Large Language Model-based agents) collaborating with humans on solving such problems is being actively explored. While the abilities of LLMs can be improved by, for example, fine-tuning, hand-crafted system prompts and scaffolding with external tools, LLMs lack endogenous mechanisms to develop expertise through experience in such settings. This work address this gap with Dialectica, a framework where agents engage in structured dialogue on defined topics, augmented by memory, self-reflection, and policy-constrained context editing. Formally, discussion is viewed as an implicit meta-reinforcement learning process. The `dialogue-trained' agents are evaluated post-hoc using judged pairwise comparisons of elicited responses. Across two model architectures (locally run Qwen3:30b and OpenAI's o4-mini) results show that enabling reflection-based context editing during discussion produces agents which dominate their baseline counterparts on Elo scores, normalized Bradley-Terry-Davidson ability, and AlphaRank mass. The predicted signatures of learning are observed qualitatively in statement and reflection logs, where reflections identify weaknesses and reliably shape subsequent statements. Agreement between quantitative and qualitative evidence supports dialogue-driven context evolution as a practical path to targeted expertise amplification in open non-verifiable domains.
Neural theorem proving has advanced rapidly in the past year, reaching IMO gold-medalist capabilities and producing formal proofs that span thousands of lines. Although such proofs are mechanically verified by formal systems like Lean, their excessive length renders them difficult for humans to comprehend and limits their usefulness for mathematical insight. Proof simplification is therefore a critical bottleneck. Yet, training data for this task is scarce, and existing methods -- mainly agentic scaffolding with off-the-shelf LLMs -- struggle with the extremely long proofs generated by RL-trained provers. We introduce ProofOptimizer, the first language model trained to simplify Lean proofs without requiring additional human supervision. ProofOptimizer is trained via expert iteration and reinforcement learning, using Lean to verify simplifications and provide training signal. At inference time, it operates within an iterative proof-shortening workflow, progressively reducing proof length. Experiments show that ProofOptimizer substantially compresses proofs generated by state-of-the-art RL-trained provers on standard benchmarks, reducing proof length by 87% on miniF2F, 57% on PutnamBench, and 49% on Seed-Prover's IMO 2025 proofs. Beyond conciseness, the simplified proofs check faster in Lean and further improve downstream prover performance when reused as training data for supervised finetuning.
We present SQuAI (https://squai.scads.ai/), a scalable and trustworthy multi-agent retrieval-augmented generation (RAG) framework for scientific question answering (QA) with large language models (LLMs). SQuAI addresses key limitations of existing RAG systems in the scholarly domain, where complex, open-domain questions demand accurate answers, explicit claims with citations, and retrieval across millions of scientific documents. Built on over 2.3 million full-text papers from arXiv.org, SQuAI employs four collaborative agents to decompose complex questions into sub-questions, retrieve targeted evidence via hybrid sparse-dense retrieval, and adaptively filter documents to improve contextual relevance. To ensure faithfulness and traceability, SQuAI integrates in-line citations for each generated claim and provides supporting sentences from the source documents. Our system improves faithfulness, answer relevance, and contextual relevance by up to +0.088 (12%) over a strong RAG baseline. We further release a benchmark of 1,000 scientific question-answer-evidence triplets to support reproducibility. With transparent reasoning, verifiable citations, and domain-wide scalability, SQuAI demonstrates how multi-agent RAG enables more trustworthy scientific QA with LLMs.
Creative services teams increasingly rely on large language models (LLMs) to accelerate ideation, yet production systems often converge on homogeneous outputs that fail to meet brand or artistic expectations. Art of X developed persona-conditioned LLM agents -- internally branded as "Sparks" and instantiated through a library of role-inspired system prompts -- to intentionally diversify agent behaviour within a multi-agent workflow. This white paper documents the problem framing, experimental design, and quantitative evidence behind the Spark agent programme. Using an LLM-as-a-judge protocol calibrated against human gold standards, we observe a mean diversity gain of +4.1 points (on a 1-10 scale) when persona-conditioned Spark agents replace a uniform system prompt, narrowing the gap to human experts to 1.0 point. We also surface evaluator bias and procedural considerations for future deployments.
The instruction-following capabilities of large language models (LLMs) are pivotal for numerous applications, from conversational agents to complex reasoning systems. However, current evaluations predominantly focus on English models, neglecting the linguistic and cultural nuances of other languages. Specifically, Korean, with its distinct syntax, rich morphological features, honorific system, and dual numbering systems, lacks a dedicated benchmark for assessing open-ended instruction-following capabilities. To address this gap, we introduce the Korean Instruction-following Task Evaluation (KITE), a comprehensive benchmark designed to evaluate both general and Korean-specific instructions. Unlike existing Korean benchmarks that focus mainly on factual knowledge or multiple-choice testing, KITE directly targets diverse, open-ended instruction-following tasks. Our evaluation pipeline combines automated metrics with human assessments, revealing performance disparities across models and providing deeper insights into their strengths and weaknesses. By publicly releasing the KITE dataset and code, we aim to foster further research on culturally and linguistically inclusive LLM development and inspire similar endeavors for other underrepresented languages.
Reinforcement learning (RL) has been pivotal in enhancing the reasoning capabilities of large language models (LLMs), but it often suffers from limited exploration and entropy collapse, where models exploit a narrow set of solutions, leading to a loss of sampling diversity and subsequently preventing RL from further improving performance. This issue is exacerbated in parallel sampling methods, where multiple outputs are drawn from the same distribution, potentially causing the model to converge to similar solutions. We propose SESA, a novel SEquential SAmpling framework that mitigates this challenge by generating diverse solution sketches sequentially before expanding them into full reasoning paths. This approach ensures broader exploration by conditioning each new output on previous ones, promoting diversity throughout the process and preventing policy collapse. Our experiments on a synthetic task show that sequential sampling consistently outperforms traditional RL methods in terms of path diversity and recovery from collapse. Further evaluations on real-world tasks demonstrate that SESA improves both the exploration of valid strategies and the overall performance of LLMs. On three agent benchmarks, SESA lifts success rates by $+0.25$, $+0.42$, and $+0.07$ absolute over the base model (up to an additional $211\%$ relative improvement over baseline RL), underscoring its exploration advantage. This work introduces a structured approach to exploration, paving the way for more effective and diverse reasoning in RL-trained LLMs. Our code is released at https://github.com/MuLabPKU/sesa.
Mobile agents rely on Large Language Models (LLMs) to plan and execute tasks on smartphone user interfaces (UIs). While cloud-based LLMs achieve high task accuracy, they require uploading the full UI state at every step, exposing unnecessary and often irrelevant information. In contrast, local LLMs avoid UI uploads but suffer from limited capacity, resulting in lower task success rates. We propose $\textbf{CORE}$, a $\textbf{CO}$llaborative framework that combines the strengths of cloud and local LLMs to $\textbf{R}$educe UI $\textbf{E}$xposure, while maintaining task accuracy for mobile agents. CORE comprises three key components: (1) $\textbf{Layout-aware block partitioning}$, which groups semantically related UI elements based on the XML screen hierarchy; (2) $\textbf{Co-planning}$, where local and cloud LLMs collaboratively identify the current sub-task; and (3) $\textbf{Co-decision-making}$, where the local LLM ranks relevant UI blocks, and the cloud LLM selects specific UI elements within the top-ranked block. CORE further introduces a multi-round accumulation mechanism to mitigate local misjudgment or limited context. Experiments across diverse mobile apps and tasks show that CORE reduces UI exposure by up to 55.6% while maintaining task success rates slightly below cloud-only agents, effectively mitigating unnecessary privacy exposure to the cloud. The code is available at https://github.com/Entropy-Fighter/CORE.
Long-form video reasoning remains a major challenge for Video Large Language Models (Video LLMs), as static uniform frame sampling leads to information dilution and obscures critical evidence. Furthermore, existing pixel-space video reasoning agents, which are designed to actively interact with the video to acquire new visual information, remain suboptimal due to their lack of rigorous reward mechanisms to enforce evidence purity and their inability to perform temporal information supplementation beyond pre-sampled frames. To address this critical gap, we propose a novel evidence-prioritized adaptive framework built upon our core philosophy: "Select Less, Reason More." Our core contribution is the evidence-aware reinforcement learning (EARL) framework, which transforms the model into an active interrogator of evidence. EARL is precisely engineered to dynamically select the most relevant frames and, crucially, to perform localized re-sampling around the selected key frames to access fine-grained temporal detail. Extensive experiments on five demanding video reasoning benchmarks demonstrate that our EARL-trained model achieves new state-of-the-art among open-source Video LLMs, simultaneously learning an effective and high-purity visual evidence selection policy. Impressively, our 7B model achieves 59.8% on LongVideoBench, 69.0% on MVBench and 64.9% on VideoMME. These results highlight the importance of prioritizing evidence purity and the effectiveness of our framework.
We present Adaptive Minds, an agentic system that treats LoRA adapters as domain-specific tools. Instead of relying on a single fine-tuned model or rigid rule-based routing, our approach empowers the base LLM itself to act as a semantic router analyzing each query and dynamically selecting the most relevant LoRA tool. This enables the agent to seamlessly switch between different domain experts on demand. By combining the flexibility of multi-agent orchestration with the efficiency of parameter-efficient fine-tuning, Adaptive Minds delivers accurate, specialized responses while preserving conversational ability. The system is built with LangGraph for workflow management, supports both API and web interfaces, and is fully open source, providing a scalable and extensible foundation for domain-adaptive AI assistance.
Developing Large Language Models (LLMs) to cooperate and compete effectively within multi-agent systems is a critical step towards more advanced intelligence. While reinforcement learning (RL) has proven effective for enhancing reasoning in single-agent tasks, its extension to multi-turn, multi-agent scenarios remains underexplored due to the challenges of long-horizon credit assignment and agent-specific advantage estimation. To address these challenges, we introduce MARS, an end-to-end RL framework that incentivizes Multi-Agent Reasoning of LLMs through Self-play in both cooperative and competitive games. MARS features a turn-level advantage estimator that aligns learning signals with each interaction for credit assignment, and an agent-specific advantage normalization to stabilize multi-agent training. By learning with self-play across cooperative and competitive games, the MARS agent trained from Qwen3-4B develops strong strategic abilities that generalize to held-out games with up to 28.7% performance improvements. More importantly, the capability acquired through self-play generalizes beyond games, yielding consistent performance gains of multi-agent systems in reasoning benchmarks. When integrated into leading multi-agent systems, our MARS agent achieves significant performance gains of 10.0% on AIME and 12.5% on GPQA-Diamond. These results establish end-to-end RL training with self-play in strategic games as a powerful approach for developing generalizable multi-agent reasoning capabilities in LLMs. Our code and models are publicly available at https://github.com/thu-nics/MARS.
Enhancing on-device large language models (LLMs) with contextual information from local data enables personalized and task-aware generation, powering use cases such as intelligent assistants and UI agents. While recent developments in neural processors have substantially improved the efficiency of prefill on mobile devices, the token-by-token generation process still suffers from high latency and limited hardware utilization due to its inherently memory-bound characteristics. This work presents CoordGen, a mobile inference framework that integrates speculative decoding with dynamic hardware scheduling to accelerate context-aware text generation on mobile devices. The framework introduces three synergistic components: (1) adaptive execution scheduling, which dynamically balances compute graphs between prefill and decoding phases; (2) context-aligned drafting, which improves speculative efficiency through lightweight online calibration to current tasks; and (3) hardware-efficient draft extension, which reuses and expands intermediate sequences to improve processing parallelism and reduce verification cost. Experiments on multiple smartphones and representative workloads show consistent improvements of up to 3.8x in generation speed and 4.7x in energy efficiency compared with existing mobile inference solutions. Component-level analysis further validates the contribution of each optimization.
Witnessed by the recent advancements on leveraging LLM for coding and multimodal understanding, we present WebGen-V, a new benchmark and framework for instruction-to-HTML generation that enhances both data quality and evaluation granularity. WebGen-V contributes three key innovations: (1) an unbounded and extensible agentic crawling framework that continuously collects real-world webpages and can leveraged to augment existing benchmarks; (2) a structured, section-wise data representation that integrates metadata, localized UI screenshots, and JSON-formatted text and image assets, explicit alignment between content, layout, and visual components for detailed multimodal supervision; and (3) a section-level multimodal evaluation protocol aligning text, layout, and visuals for high-granularity assessment. Experiments with state-of-the-art LLMs and ablation studies validate the effectiveness of our structured data and section-wise evaluation, as well as the contribution of each component. To the best of our knowledge, WebGen-V is the first work to enable high-granularity agentic crawling and evaluation for instruction-to-HTML generation, providing a unified pipeline from real-world data acquisition and webpage generation to structured multimodal assessment.
Large Language Models (LLMs) as interactive agents show significant promise in Knowledge Graph Question Answering (KGQA) but often struggle with the semantic gap between natural language queries and structured knowledge graph (KG) representations. This leads to suboptimal planning and inefficient exploration on KG, while training-free approaches often underutilize valuable reasoning patterns in training data. To address these limitations, we propose a novel framework, Exemplar-Guided Planning (EGP), which enhances the planning capabilities of LLM agents for KGQA. EGP first preprocesses the training set questions via entity templating to normalize semantic variations. It then retrieves highly similar exemplary questions and their successful reasoning paths from this preprocessed set using semantic embeddings and an efficient FAISS index. These retrieved exemplars dynamically guide the LLM's planning process in two key phases: (1) Task Decomposition, by aligning generated sub-objectives with proven reasoning steps, and (2) Relation Exploration, by providing high-quality auxiliary information to improve relation pruning accuracy. Additionally, we introduce a Smart Lookahead mechanism during relation exploration to improve efficiency by preemptively exploring promising paths and potentially terminating exploration earlier. We apply EGP to the Plan-on-Graph (PoG) framework, termed PoG-EGP. Extensive experiments on two real-world KGQA datasets, WebQSP and CWQ, demonstrate that PoG-EGP significantly improves over the baseline PoG system and other compared methods.
Riding on the success of LLMs with retrieval-augmented generation (RAG), there has been a growing interest in augmenting agent systems with external memory databases. However, the existing systems focus on storing text information in their memory, ignoring the importance of multimodal signals. Motivated by the multimodal nature of human memory, we present AUGUSTUS, a multimodal agent system aligned with the ideas of human memory in cognitive science. Technically, our system consists of 4 stages connected in a loop: (i) encode: understanding the inputs; (ii) store in memory: saving important information; (iii) retrieve: searching for relevant context from memory; and (iv) act: perform the task. Unlike existing systems that use vector databases, we propose conceptualizing information into semantic tags and associating the tags with their context to store them in a graph-structured multimodal contextual memory for efficient concept-driven retrieval. Our system outperforms the traditional multimodal RAG approach while being 3.5 times faster for ImageNet classification and outperforming MemGPT on the MSC benchmark.
Most existing software lacks accessible Application Programming Interfaces (APIs), requiring agents to operate solely through pixel-based Graphical User Interfaces (GUIs). In this API-free setting, large language model (LLM)-based agents face severe efficiency bottlenecks: limited to local visual experiences, they make myopic decisions and rely on inefficient trial-and-error, hindering both skill acquisition and long-term planning. To address these challenges, we propose KG-Agent, an experience-driven learning framework that structures an agent's raw pixel-level interactions into a persistent State-Action Knowledge Graph (SA-KG). KG-Agent overcomes inefficient exploration by linking functionally similar but visually distinct GUI states, forming a rich neighborhood of experience that enables the agent to generalize from a diverse set of historical strategies. To support long-horizon reasoning, we design a hybrid intrinsic reward mechanism based on the graph topology, combining a state value reward for exploiting known high-value pathways with a novelty reward that encourages targeted exploration. This approach decouples strategic planning from pure discovery, allowing the agent to effectively value setup actions with delayed gratification. We evaluate KG-Agent in two complex, open-ended GUI-based decision-making environments (Civilization V and Slay the Spire), demonstrating significant improvements in exploration efficiency and strategic depth over the state-of-the-art methods.
In the current era of big data, extracting deep insights from massive, heterogeneous, and complexly associated multi-dimensional data has become a significant challenge. Large Language Models (LLMs) perform well in natural language understanding and generation, but still suffer from "hallucination" issues when processing structured knowledge and are difficult to update in real-time. Although Knowledge Graphs (KGs) can explicitly store structured knowledge, their static nature limits dynamic interaction and analytical capabilities. Therefore, this paper proposes a multi-dimensional data analysis method based on the interactions between LLM agents and KGs, constructing a dynamic, collaborative analytical ecosystem. This method utilizes LLM agents to automatically extract product data from unstructured data, constructs and visualizes the KG in real-time, and supports users in deep exploration and analysis of graph nodes through an interactive platform. Experimental results show that this method has significant advantages in product ecosystem analysis, relationship mining, and user-driven exploratory analysis, providing new ideas and tools for multi-dimensional data analysis.
Document understanding is critical for applications from financial analysis to scientific discovery. Current approaches, whether OCR-based pipelines feeding Large Language Models (LLMs) or native Multimodal LLMs (MLLMs), face key limitations: the former loses structural detail, while the latter struggles with context modeling. Retrieval-Augmented Generation (RAG) helps ground models in external data, but documents' multimodal nature, i.e., combining text, tables, charts, and layout, demands a more advanced paradigm: Multimodal RAG. This approach enables holistic retrieval and reasoning across all modalities, unlocking comprehensive document intelligence. Recognizing its importance, this paper presents a systematic survey of Multimodal RAG for document understanding. We propose a taxonomy based on domain, retrieval modality, and granularity, and review advances involving graph structures and agentic frameworks. We also summarize key datasets, benchmarks, and applications, and highlight open challenges in efficiency, fine-grained representation, and robustness, providing a roadmap for future progress in document AI.
Critical reading is a primary way through which researchers develop their critical thinking skills. While exchanging thoughts and opinions with peers can strengthen critical reading, junior researchers often lack access to peers who can offer diverse perspectives. To address this gap, we designed an in-situ thought exchange interface informed by peer feedback from a formative study (N=8) to support junior researchers' critical paper reading. We evaluated the effects of thought exchanges under three conditions (no-agent, single-agent, and multi-agent) with 46 junior researchers over two weeks. Our results showed that incorporating agent-mediated thought exchanges during paper reading significantly improved participants' critical thinking scores compared to the no-agent condition. In the single-agent condition, participants more frequently made reflective annotations on the paper content. In the multi-agent condition, participants engaged more actively with agents' responses. Our qualitative analysis further revealed that participants compared and analyzed multiple perspectives in the multi-agent condition. This work contributes to understanding in-situ AI-based support for critical paper reading through thought exchanges and offers design implications for future research.
A core challenge for autonomous LLM agents in collaborative settings is balancing robust privacy understanding and preservation alongside task efficacy. Existing privacy benchmarks only focus on simplistic, single-turn interactions where private information can be trivially omitted without affecting task outcomes. In this paper, we introduce MAGPIE (Multi-AGent contextual PrIvacy Evaluation), a novel benchmark of 200 high-stakes tasks designed to evaluate privacy understanding and preservation in multi-agent collaborative, non-adversarial scenarios. MAGPIE integrates private information as essential for task resolution, forcing agents to balance effective collaboration with strategic information control. Our evaluation reveals that state-of-the-art agents, including GPT-5 and Gemini 2.5-Pro, exhibit significant privacy leakage, with Gemini 2.5-Pro leaking up to 50.7% and GPT-5 up to 35.1% of the sensitive information even when explicitly instructed not to. Moreover, these agents struggle to achieve consensus or task completion and often resort to undesirable behaviors such as manipulation and power-seeking (e.g., Gemini 2.5-Pro demonstrating manipulation in 38.2% of the cases). These findings underscore that current LLM agents lack robust privacy understanding and are not yet adequately aligned to simultaneously preserve privacy and maintain effective collaboration in complex environments.
Reinforcement Learning (RL) enables agents to learn optimal decision-making strategies through interaction with an environment, yet training from scratch on complex tasks can be highly inefficient. Transfer learning (TL), widely successful in large language models (LLMs), offers a promising direction for enhancing RL efficiency by leveraging pre-trained models. This paper investigates policy transfer, a TL approach that initializes learning in a target RL task using a policy from a related source task, in the context of continuous-time linear quadratic regulators (LQRs) with entropy regularization. We provide the first theoretical proof of policy transfer for continuous-time RL, proving that a policy optimal for one LQR serves as a near-optimal initialization for closely related LQRs, while preserving the original algorithm's convergence rate. Furthermore, we introduce a novel policy learning algorithm for continuous-time LQRs that achieves global linear and local super-linear convergence. Our results demonstrate both theoretical guarantees and algorithmic benefits of transfer learning in continuous-time RL, addressing a gap in existing literature and extending prior work from discrete to continuous time settings. As a byproduct of our analysis, we derive the stability of a class of continuous-time score-based diffusion models via their connection with LQRs.
Simulating human reasoning in open-ended tasks has been a long-standing aspiration in AI and cognitive science. While large language models now approximate human responses at scale, they remain tuned to population-level consensus, often erasing the individuality of reasoning styles and belief trajectories. To advance the vision of more human-like reasoning in machines, we introduce HugAgent (Human-Grounded Agent Benchmark), a benchmark for average-to-individual reasoning adaptation. The task is to predict how a specific person would reason and update their beliefs in novel scenarios, given partial evidence of their past views. HugAgent adopts a dual-track design: a synthetic track for scale and systematic stress tests, and a human track for ecologically valid, "out-loud" reasoning data. This design enables scalable, reproducible evaluation of intra-agent fidelity: whether models can capture not just what people believe, but how their reasoning evolves. Experiments with state-of-the-art LLMs reveal persistent adaptation gaps, positioning HugAgent as the first extensible benchmark for aligning machine reasoning with the individuality of human thought. Our benchmark and chatbot are open-sourced as HugAgent (https://anonymous.4open.science/r/HugAgent) and TraceYourThinking (https://anonymous.4open.science/r/trace-your-thinking).
Large Language Models (LLMs) as agents often struggle in out-of-distribution (OOD) scenarios. Real-world environments are complex and dynamic, governed by task-specific rules and stochasticity, which makes it difficult for LLMs to ground their internal knowledge in those dynamics. Under such OOD conditions, vanilla RL training often fails to scale; we observe Pass@k--the probability that at least one of (k) sampled trajectories succeeds--drops markedly across training steps, indicating brittle exploration and limited generalization. Inspired by model-based reinforcement learning, we hypothesize that equipping LLM agents with an internal world model can better align reasoning with environmental dynamics and improve decision-making. We show how to encode this world model by decomposing it into two components: state representation and transition modeling. Building on this, we introduce SPA, a simple reinforcement learning framework that cold-starts the policy via a Self-Play supervised finetuning (SFT) stage to learn the world model by interacting with the environment, then uses it to simulate future states prior to policy optimization. This simple initialization outperforms the online world-modeling baseline and greatly boosts the RL-based agent training performance. Experiments across diverse environments like Sokoban, FrozenLake, and Sudoku show that our approach significantly improves performance. For example, SPA boosts the Sokoban success rate from 25.6% to 59.8% and raises the FrozenLake score from 22.1% to 70.9% for the Qwen2.5-1.5B-Instruct model.
The design of complex machines stands as both a marker of human intelligence and a foundation of engineering practice. Given recent advances in large language models (LLMs), we ask whether they, too, can learn to create. We approach this question through the lens of compositional machine design: a task in which machines are assembled from standardized components to meet functional demands like locomotion or manipulation in a simulated physical environment. To support this investigation, we introduce BesiegeField, a testbed built on the machine-building game Besiege, which enables part-based construction, physical simulation and reward-driven evaluation. Using BesiegeField, we benchmark state-of-the-art LLMs with agentic workflows and identify key capabilities required for success, including spatial reasoning, strategic assembly, and instruction-following. As current open-source models fall short, we explore reinforcement learning (RL) as a path to improvement: we curate a cold-start dataset, conduct RL finetuning experiments, and highlight open challenges at the intersection of language, machine design, and physical reasoning.
Digital agents require diverse, large-scale UI trajectories to generalize across real-world tasks, yet collecting such data is prohibitively expensive in both human annotation, infra and engineering perspectives. To this end, we introduce $\textbf{UI-Simulator}$, a scalable paradigm that generates structured UI states and transitions to synthesize training trajectories at scale. Our paradigm integrates a digital world simulator for diverse UI states, a guided rollout process for coherent exploration, and a trajectory wrapper that produces high-quality and diverse trajectories for agent training. We further propose $\textbf{UI-Simulator-Grow}$, a targeted scaling strategy that enables more rapid and data-efficient scaling by prioritizing high-impact tasks and synthesizes informative trajectory variants. Experiments on WebArena and AndroidWorld show that UI-Simulator rivals or surpasses open-source agents trained on real UIs with significantly better robustness, despite using weaker teacher models. Moreover, UI-Simulator-Grow matches the performance of Llama-3-70B-Instruct using only Llama-3-8B-Instruct as the base model, highlighting the potential of targeted synthesis scaling paradigm to continuously and efficiently enhance the digital agents.
Large language model (LLM)-based agents are increasingly trained with reinforcement learning (RL) to enhance their ability to interact with external environments through tool use, particularly in search-based settings that require multi-turn reasoning and knowledge acquisition. However, existing approaches typically rely on outcome-based rewards that are only provided at the final answer. This reward sparsity becomes particularly problematic in multi-turn settings, where long trajectories exacerbate two critical issues: (i) advantage collapse, where all rollouts receive identical rewards and provide no useful learning signals, and (ii) lack of fine-grained credit assignment, where dependencies between turns are obscured, especially in long-horizon tasks. In this paper, we propose Information Gain-based Policy Optimization (IGPO), a simple yet effective RL framework that provides dense and intrinsic supervision for multi-turn agent training. IGPO models each interaction turn as an incremental process of acquiring information about the ground truth, and defines turn-level rewards as the marginal increase in the policy's probability of producing the correct answer. Unlike prior process-level reward approaches that depend on external reward models or costly Monte Carlo estimation, IGPO derives intrinsic rewards directly from the model's own belief updates. These intrinsic turn-level rewards are combined with outcome-level supervision to form dense reward trajectories. Extensive experiments on both in-domain and out-of-domain benchmarks demonstrate that IGPO consistently outperforms strong baselines in multi-turn scenarios, achieving higher accuracy and improved sample efficiency.
Pairwise comparisons of large language models using total variation distance mutual information (TVD-MI) produce binary critic decisions per pair. We show that averaging TVD-MI's binary trials yields centered-probability scores with additive structure suitable for item-response theory (IRT) without nonlinear link functions. Maximum-likelihood approaches to IRT use logistic links, but we find empirically that these transformations introduce curvature that breaks additivity: across three domains, the identity link yields median curl on raw data of 0.080-0.150 (P95 = [0.474, 0.580]), whereas probit/logit introduce substantially higher violations (median [0.245, 0.588], P95 [0.825, 2.252]). We derive this clipped-linear model from Gini entropy maximization, yielding a box-constrained least-squares formulation that handles boundary saturation. At 33% coverage, we achieve holdout RMSE $0.117 \pm 0.008$ while preserving agent rankings (Spearman $\rho = 0.972 \pm 0.015$), three times fewer evaluations than full dense. Judge robustness analysis (GPT-4o-mini vs. Llama3-70b) shows strong agreement in agent rankings ($\rho = 0.872$) and consistent identity-link advantage. TVD-MI's geometry is best preserved by identity mapping for efficient LLM evaluation, applicable to other bounded-response domains.
The Enterprise Intelligence Platform must integrate logs from numerous third-party vendors in order to perform various downstream tasks. However, vendor documentation is often unavailable at test time. It is either misplaced, mismatched, poorly formatted, or incomplete, which makes schema mapping challenging. We introduce a reinforcement learning agent that can self-improve without labeled examples or model weight updates. During inference, the agent: 1) Identifies ambiguous field-mapping attempts. 2) Generates targeted web-search queries to gather external evidence. 3) Applies a confidence-based reward to iteratively refine its mappings. To demonstrate this concept, we converted Microsoft Defender for Endpoint logs into a common schema. Our method increased mapping accuracy from 56.4\%(LLM-only) to 72.73\%(RAG) to 93.94\% over 100 iterations using GPT-4o. At the same time, it reduced the number of low-confidence mappings requiring expert review by 85\%. This new approach provides an evidence-driven, transparent method for solving future industry problems, paving the way for more robust, accountable, scalable, efficient, flexible, adaptable, and collaborative solutions.