Existing appliance assets suffer from poor rendering, incomplete mechanisms, and misalignment with manuals, leading to simulation-reality gaps that hinder appliance manipulation development. In this work, we introduce the RealAppliance dataset, comprising 100 high-fidelity appliances with complete physical, electronic mechanisms, and program logic aligned with their manuals. Based on these assets, we propose the RealAppliance-Bench benchmark, which evaluates multimodal large language models and embodied manipulation planning models across key tasks in appliance manipulation planning: manual page retrieval, appliance part grounding, open-loop manipulation planning, and closed-loop planning adjustment. Our analysis of model performances on RealAppliance-Bench provides insights for advancing appliance manipulation research.

We study idiom-based visual puns--images that align an idiom's literal and figurative meanings--and present an iterative framework that coordinates a large language model (LLM), a text-to-image model (T2IM), and a multimodal LLM (MLLM) for automatic generation and evaluation. Given an idiom, the system iteratively (i) generates detailed visual prompts, (ii) synthesizes an image, (iii) infers the idiom from the image, and (iv) refines the prompt until recognition succeeds or a step limit is reached. Using 1,000 idioms as inputs, we synthesize a corresponding dataset of visual pun images with paired prompts, enabling benchmarking of both generation and understanding. Experiments across 10 LLMs, 10 MLLMs, and one T2IM (Qwen-Image) show that MLLM choice is the primary performance driver: GPT achieves the highest accuracies, Gemini follows, and the best open-source MLLM (Gemma) is competitive with some closed models. On the LLM side, Claude attains the strongest average performance for prompt generation.

However, most existing multimodal mathematical reasoning benchmarks focus primarily on 2D plane geometry, rely on static datasets prone to data contamination and memorization, and evaluate models solely by final answers, overlooking the reasoning process. T o address these limitations, we introduce DynaSolidGeo, the first dynamic benchmark for evaluating genuine spatial reasoning in Vision-Language Models (VLMs). Constructed through a semi-automatic annotation pipeline, DynaSolidGeo contains 503 expert-curated seed questions that can, in principle, dynamically generate an unbounded number of diverse multimodal text-visual instances. Beyond answer accuracy, we incorporate process evaluation based on expert-annotated reasoning chains to measure logical validity and causal coherence. Experiments across representative open-source and closed-source VLMs reveal large performance gaps, severe degradation in dynamic settings, and poor performance on tasks requiring high-level spatial intelligence, such as mental rotation and visualization. The code and dataset are available at DynaSolidGeo.

Large language models (LLMs) are becoming increasingly widespread. Organizations that want to use AI for productivity now face an important decision. They can subscribe to commercial LLM services or deploy models on their own infrastructure. Cloud services from providers such as OpenAI, Anthropic, and Google are attractive because they provide easy access to state-of-the-art models and are easy to scale. However, concerns about data privacy, the difficulty of switching service providers, and long-term operating costs have driven interest in local deployment of open-source models. This paper presents a cost-benefit analysis framework to help organizations determine when on-premise LLM deployment becomes economically viable compared to commercial subscription services. We consider the hardware requirements, operational expenses, and performance benchmarks of the latest open-source models, including Qwen, Llama, Mistral, and etc. Then we compare the total cost of deploying these models locally with the major cloud providers subscription fee. Our findings provide an estimated breakeven point based on usage levels and performance needs. These results give organizations a practical framework for planning their LLM strategies.

Large language model (LLM) agents have exhibited strong problem-solving competence across domains like research and coding. Yet, it remains underexplored whether LLM agents can tackle compounding real-world problems that require a diverse set of tools to complete. Given a broad, heterogeneous tool repository, LLM agents must not only select appropriate tools based on task planning analysis but also strategically schedule the execution order to ensure efficiency. This paper introduces TPS-Bench to benchmark the ability of LLM agents in solving such problems that demand Tool Planning and Scheduling. TPS-Bench collects 200 compounding tasks of two difficulty levels, based on a tool repository containing hundreds of model context protocol (MCP) tools. In particular, each task is composed of multiple subtasks, such as web search, map navigation, calendar checking, etc., and each subtask can be completed by a basic tool. Our evaluation emphasizes both task completion rate and efficiency. The empirical studies on popular closed-source and open-source LLMs indicate that most models can perform reasonable tool planning, but differ in scheduling. For example, GLM-4.5 achieves an outperforming task completion rate of 64.72% with extensive sequential tool calls, hence suffering from significantly long execution time. By contrast, GPT-4o prioritizes parallel tool calls but achieves only a 45.08% completion rate. Considering reinforcement learning (RL) can be a viable way to improve the scheduling efficiency without compromising performance, we perform an initial study on Qwen3-1.7B and witness a 14% reduction in execution time alongside a 6% gain in task completion rate based on rarely 100 RL training samples. Our code is available https://github.com/hanwenxu1/mcp-agent.

Large Language Model (LLM) agents have recently shown strong potential in domains such as automated coding, deep research, and graphical user interface manipulation. However, training them to succeed on long-horizon, domain-specialized tasks remains challenging. Current methods primarily fall into two categories. The first relies on dense human annotations through behavior cloning, which is prohibitively expensive for long-horizon tasks that can take days or months. The second depends on outcome-driven sampling, which often collapses due to the rarity of valid positive trajectories on domain-specialized tasks. We introduce Apollo, a sampling framework that integrates asynchronous human guidance with action-level data filtering. Instead of requiring annotators to shadow every step, Apollo allows them to intervene only when the agent drifts from a promising trajectory, by providing prior knowledge, strategic advice, etc. This lightweight design makes it possible to sustain interactions for over 30 hours and produces valuable trajectories at a lower cost. Apollo then applies supervision control to filter out sub-optimal actions and prevent error propagation. Together, these components enable reliable and effective data collection in long-horizon environments. To demonstrate the effectiveness of Apollo, we evaluate it using InnovatorBench. Our experiments show that when applied to train the GLM-4.5 model on InnovatorBench, Apollo achieves more than a 50% improvement over the untrained baseline and a 28% improvement over a variant trained without human interaction. These results highlight the critical role of human-in-the-loop sampling and the robustness of Apollo's design in handling long-horizon, domain-specialized tasks.

Large Multimodal Models (LMMs) are increasingly capable of answering medical questions that require joint reasoning over images and text, yet training general medical VQA systems is impeded by the lack of large, openly usable, high-quality corpora. We present MedVLSynther, a rubric-guided generator-verifier framework that synthesizes high-quality multiple-choice VQA items directly from open biomedical literature by conditioning on figures, captions, and in-text references. The generator produces self-contained stems and parallel, mutually exclusive options under a machine-checkable JSON schema; a multi-stage verifier enforces essential gates (self-containment, single correct answer, clinical validity, image-text consistency), awards fine-grained positive points, and penalizes common failure modes before acceptance. Applying this pipeline to PubMed Central yields MedSynVQA: 13,087 audited questions over 14,803 images spanning 13 imaging modalities and 28 anatomical regions. Training open-weight LMMs with reinforcement learning using verifiable rewards improves accuracy across six medical VQA benchmarks, achieving averages of 55.85 (3B) and 58.15 (7B), with up to 77.57 on VQA-RAD and 67.76 on PathVQA, outperforming strong medical LMMs. A Ablations verify that both generation and verification are necessary and that more verified data consistently helps, and a targeted contamination analysis detects no leakage from evaluation suites. By operating entirely on open literature and open-weight models, MedVLSynther offers an auditable, reproducible, and privacy-preserving path to scalable medical VQA training data.

While Large Language Models (LLMs) are increasingly envisioned as intelligent assistants for personalized learning, systematic head-to-head evaluations in authentic learning scenarios remain scarce. This study presents an empirical comparison of three state-of-the-art LLMs on a tutoring task simulating a realistic learning setting. Using a dataset containing a student's responses to ten mixed-format questions with correctness labels, each model was asked to (i) analyze the quiz to identify underlying knowledge components, (ii) infer the student's mastery profile, and (iii) generate targeted guidance for improvement. To mitigate subjectivity and evaluator bias, Gemini was employed as a virtual judge to perform pairwise comparisons across multiple dimensions: accuracy, clarity, actionability, and appropriateness. Results analyzed via the Bradley-Terry model reveal that GPT-4o is generally preferred, producing feedback that is more informative and better structured than its counterparts, whereas DeepSeek-V3 and GLM-4.5 demonstrate intermittent strengths but lower consistency. These findings highlight the feasibility of deploying LLMs as advanced teaching assistants for individualized support and provide methodological insights for subsequent empirical research on LLM-driven personalized learning.

The growing deployment of large language model (LLM) based agents that interact with external environments has created new attack surfaces for adversarial manipulation. One major threat is indirect prompt injection, where attackers embed malicious instructions in external environment output, causing agents to interpret and execute them as if they were legitimate prompts. While previous research has focused primarily on plain-text injection attacks, we find a significant yet underexplored vulnerability: LLMs' dependence on structured chat templates and their susceptibility to contextual manipulation through persuasive multi-turn dialogues. To this end, we introduce ChatInject, an attack that formats malicious payloads to mimic native chat templates, thereby exploiting the model's inherent instruction-following tendencies. Building on this foundation, we develop a persuasion-driven Multi-turn variant that primes the agent across conversational turns to accept and execute otherwise suspicious actions. Through comprehensive experiments across frontier LLMs, we demonstrate three critical findings: (1) ChatInject achieves significantly higher average attack success rates than traditional prompt injection methods, improving from 5.18% to 32.05% on AgentDojo and from 15.13% to 45.90% on InjecAgent, with multi-turn dialogues showing particularly strong performance at average 52.33% success rate on InjecAgent, (2) chat-template-based payloads demonstrate strong transferability across models and remain effective even against closed-source LLMs, despite their unknown template structures, and (3) existing prompt-based defenses are largely ineffective against this attack approach, especially against Multi-turn variants. These findings highlight vulnerabilities in current agent systems.

We define Agency as the emergent capacity of AI systems to function as autonomous agents actively discovering problems, formulating hypotheses, and executing solutions through self-directed engagement with environments and tools. This fundamental capability marks the dawn of the Age of AI Agency, driven by a critical industry shift: the urgent need for AI systems that don't just think, but work. While current AI excels at reasoning and generating responses, industries demand autonomous agents that can execute tasks, operate tools, and drive real-world outcomes. As agentic intelligence becomes the defining characteristic separating cognitive systems from productive workers, efficiently cultivating machine autonomy becomes paramount. Current approaches assume that more data yields better agency, following traditional scaling laws from language modeling. We fundamentally challenge this paradigm. LIMI (Less Is More for Intelligent Agency) demonstrates that agency follows radically different development principles. Through strategic focus on collaborative software development and scientific research workflows, we show that sophisticated agentic intelligence can emerge from minimal but strategically curated demonstrations of autonomous behavior. Using only 78 carefully designed training samples, LIMI achieves 73.5% on comprehensive agency benchmarks, dramatically outperforming state-of-the-art models: Kimi-K2-Instruct (24.1%), DeepSeek-V3.1 (11.9%), Qwen3-235B-A22B-Instruct (27.5%), and GLM-4.5 (45.1%). Most strikingly, LIMI demonstrates 53.7% improvement over models trained on 10,000 samples-achieving superior agentic intelligence with 128 times fewer samples. Our findings establish the Agency Efficiency Principle: machine autonomy emerges not from data abundance but from strategic curation of high-quality agentic demonstrations.