Programmable task-oriented dialogue (TOD) agents enable language models to follow structured dialogue policies, but their effectiveness hinges on accurate state tracking. We present PyTOD, an agent that generates executable code to track dialogue state and uses policy and execution feedback for efficient error correction. To this end, PyTOD employs a simple constrained decoding approach, using a language model instead of grammar rules to follow API schemata. This leads to state-of-the-art state tracking performance on the challenging SGD benchmark. Our experiments show that PyTOD surpasses strong baselines in both accuracy and robust user goal estimation as the dialogue progresses, demonstrating the effectiveness of execution-aware state tracking.

Modern optimization algorithms that incorporate momentum and adaptive step-size offer improved performance in numerous challenging deep learning tasks. However, their effectiveness is often highly sensitive to the choice of hyperparameters, especially the step-size. Tuning these parameters is often difficult, resource-intensive, and time-consuming. Therefore, recent efforts have been directed toward enhancing the stability of optimizers across a wide range of hyperparameter choices [Schaipp et al., 2024]. In this paper, we introduce an algorithm that matches the performance of state-of-the-art optimizers while improving stability to the choice of the step-size hyperparameter through a novel adaptation of the NGN step-size method [Orvieto and Xiao, 2024]. Specifically, we propose a momentum-based version (NGN-M) that attains the standard convergence rate of $\mathcal{O}(1/\sqrt{K})$ under less restrictive assumptions, without the need for interpolation condition or assumptions of bounded stochastic gradients or iterates, in contrast to previous approaches. Additionally, we empirically demonstrate that the combination of the NGN step-size with momentum results in enhanced robustness to the choice of the step-size hyperparameter while delivering performance that is comparable to or surpasses other state-of-the-art optimizers.

LongRecall. The completeness of machine-generated text, ensuring that it captures all relevant information, is crucial in domains such as medicine and law and in tasks like list-based question answering (QA), where omissions can have serious consequences. However, existing recall metrics often depend on lexical overlap, leading to errors with unsubstantiated entities and paraphrased answers, while LLM-as-a-Judge methods with long holistic prompts capture broader semantics but remain prone to misalignment and hallucinations without structured verification. We introduce LongRecall, a general three-stage recall evaluation framework that decomposes answers into self-contained facts, successively narrows plausible candidate matches through lexical and semantic filtering, and verifies their alignment through structured entailment checks. This design reduces false positives and false negatives while accommodating diverse phrasings and contextual variations, serving as a foundational building block for systematic recall assessment. We evaluate LongRecall on three challenging long-form QA benchmarks using both human annotations and LLM-based judges, demonstrating substantial improvements in recall accuracy over strong lexical and LLM-as-a-Judge baselines.

Heliophysics is central to understanding and forecasting space weather events and solar activity. Despite decades of high-resolution observations from the Solar Dynamics Observatory (SDO), most models remain task-specific and constrained by scarce labeled data, limiting their capacity to generalize across solar phenomena. We introduce Surya, a 366M parameter foundation model for heliophysics designed to learn general-purpose solar representations from multi-instrument SDO observations, including eight Atmospheric Imaging Assembly (AIA) channels and five Helioseismic and Magnetic Imager (HMI) products. Surya employs a spatiotemporal transformer architecture with spectral gating and long--short range attention, pretrained on high-resolution solar image forecasting tasks and further optimized through autoregressive rollout tuning. Zero-shot evaluations demonstrate its ability to forecast solar dynamics and flare events, while downstream fine-tuning with parameter-efficient Low-Rank Adaptation (LoRA) shows strong performance on solar wind forecasting, active region segmentation, solar flare forecasting, and EUV spectra. Surya is the first foundation model in heliophysics that uses time advancement as a pretext task on full-resolution SDO data. Its novel architecture and performance suggest that the model is able to learn the underlying physics behind solar evolution.

Tool calling has emerged as a critical capability for AI agents to interact with the real world and solve complex tasks. While the Model Context Protocol (MCP) provides a powerful standardized framework for tool integration, there is a significant gap in benchmarking how well AI agents can effectively solve multi-step tasks using diverse MCP tools in realistic, dynamic scenarios. In this work, we present LiveMCP-101, a benchmark of 101 carefully curated real-world queries, refined through iterative LLM rewriting and manual review, that require coordinated use of multiple MCP tools including web search, file operations, mathematical reasoning, and data analysis. Moreover, we introduce a novel evaluation approach that leverages ground-truth execution plans rather than raw API outputs, better reflecting the evolving nature of real-world environments. Experiments show that even frontier LLMs achieve a success rate below 60\%, highlighting major challenges in tool orchestration. Detailed ablations and error analysis further reveal distinct failure modes and inefficiencies in token usage, pointing to concrete directions for advancing current models. LiveMCP-101 sets a rigorous standard for evaluating real-world agent capabilities, advancing toward autonomous AI systems that reliably execute complex tasks through tool use.

Large Language Models (LLMs) have made significant strides in reasoning tasks through methods like chain-of-thought (CoT) reasoning. However, they often fall short in tasks requiring precise computations. Tool-Integrated Reasoning (TIR) has emerged as a solution by incorporating external tools into the reasoning process. Nevertheless, the generalization of TIR in improving the reasoning ability of LLM is still unclear. Additionally, whether TIR has improved the model's reasoning behavior and helped the model think remains to be studied. We introduce ReasonZoo, a comprehensive benchmark encompassing nine diverse reasoning categories, to evaluate the effectiveness of TIR across various domains. Additionally, we propose two novel metrics, Performance-Aware Cost (PAC) and Area Under the Performance-Cost Curve (AUC-PCC), to assess reasoning efficiency. Our empirical evaluation demonstrates that TIR-enabled models consistently outperform their non-TIR counterparts in both mathematical and non-mathematical tasks. Furthermore, TIR enhances reasoning efficiency, as evidenced by improved PAC and AUC-PCC, indicating reduced overthinking and more streamlined reasoning. These findings underscore the domain-general benefits of TIR and its potential to advance LLM capabilities in complex reasoning tasks.

Accurate diagnosis with medical large language models is hindered by knowledge gaps and hallucinations. Retrieval and tool-augmented methods help, but their impact is limited by weak use of external knowledge and poor feedback-reasoning traceability. To address these challenges, We introduce Deep-DxSearch, an agentic RAG system trained end-to-end with reinforcement learning (RL) that enables steer tracebale retrieval-augmented reasoning for medical diagnosis. In Deep-DxSearch, we first construct a large-scale medical retrieval corpus comprising patient records and reliable medical knowledge sources to support retrieval-aware reasoning across diagnostic scenarios. More crutially, we frame the LLM as the core agent and the retrieval corpus as its environment, using tailored rewards on format, retrieval, reasoning structure, and diagnostic accuracy, thereby evolving the agentic RAG policy from large-scale data through RL. Experiments demonstrate that our end-to-end agentic RL training framework consistently outperforms prompt-engineering and training-free RAG approaches across multiple data centers. After training, Deep-DxSearch achieves substantial gains in diagnostic accuracy, surpassing strong diagnostic baselines such as GPT-4o, DeepSeek-R1, and other medical-specific frameworks for both common and rare disease diagnosis under in-distribution and out-of-distribution settings. Moreover, ablation studies on reward design and retrieval corpus components confirm their critical roles, underscoring the uniqueness and effectiveness of our approach compared with traditional implementations. Finally, case studies and interpretability analyses highlight improvements in Deep-DxSearch's diagnostic policy, providing deeper insight into its performance gains and supporting clinicians in delivering more reliable and precise preliminary diagnoses. See https://github.com/MAGIC-AI4Med/Deep-DxSearch.

This work enhances the ability of large language models (LLMs) to perform complex reasoning in 3D scenes. Recent work has addressed the 3D situated reasoning task by invoking tool usage through large language models. Large language models call tools via APIs and integrate the generated programs through a chain of thought to solve problems based on the program results. However, due to the simplicity of the questions in the dataset, the generated program reasoning chains are relatively short. To solve this main challenge, in this paper, we introduce DeepThink3D to enhance the tool usage of LLMs in complex 3D situated reasoning tasks. Our work proposes a combinatorial and iterative evolutionary approach on the SQA3D benchmark to generate more complex questions. Building on this foundation, we fine-tune the large language model to make it more proficient in using 3D tools. By employing Direct Preference Optimization (DPO), we directly optimize the toolchain strategies generated by models, thereby enhancing their accuracy in complex tasks.

With the prospect of autonomous artificial intelligence (AI) agents, studying their tendency for cooperative behavior becomes an increasingly relevant topic. This study is inspired by the super-additive cooperation theory, where the combined effects of repeated interactions and inter-group rivalry have been argued to be the cause for cooperative tendencies found in humans. We devised a virtual tournament where language model agents, grouped into teams, face each other in a Prisoner's Dilemma game. By simulating both internal team dynamics and external competition, we discovered that this blend substantially boosts both overall and initial, one-shot cooperation levels (the tendency to cooperate in one-off interactions). This research provides a novel framework for large language models to strategize and act in complex social scenarios and offers evidence for how intergroup competition can, counter-intuitively, result in more cooperative behavior. These insights are crucial for designing future multi-agent AI systems that can effectively work together and better align with human values.