The paper investigates whether large language model (LLM) agents can autonomously manage long‑horizon physical tasks without human intervention. It proposes a multi‑agent framework that combines planning, tool calling, observation, and verification, and tests it on agricultural tasks under varying weather conditions. Results show that zero‑shot LLM agents match reinforcement learning (RL) agents in the same environment and outperform RL when the environment shifts, suggesting a viable path for self‑adaptive physical AI.
By Varun Kaushik, Yayun Tan, Xiaofan Yu
The paper introduces a new approach called Homeostatic Continual Learning, designed to allow an AI agent to learn continuously in a changing environment without catastrophic forgetting. The method identifies outliers in environmental data when the agent’s output deviates, enabling the agent to incrementally refine its model and policy across increasingly diverse contexts. The authors also propose extending the method to build a world model that factorizes objects into features, abstracts them into comparable concept instances, and maps concepts to intents via features, while outlining necessary future work and broader AI connections.
By Yue Jin
DynSTEER is a dynamic stage‑wise trajectory evaluation framework designed for large language model agents performing long‑horizon tasks. It segments rollouts into stages anchored by key actions, uses a path‑tolerant milestone graph to accommodate diverse valid strategies, and adaptively routes queries to multi‑tier judges while halting unrecoverable executions early. Experiments show it improves evaluation discriminability by 85.2% over native methods, separates all model pairs with statistical significance, and saves 34.51% of execution steps on failed rollouts.
By Zhichao Shi, Wenjie Zhang, Xuhui Jiang, Xiaojun Wu, Cehao Yang, Chengjin Xu, Jian Guo, Yuanzhuo Wang
MUSE is a story‑generation engine that applies Robert McKee’s narrative theory to guide decisions about plot, character, and language throughout planning, drafting, and revision. It structures story knowledge into rule atoms, semantic consolidations, and mechanisms, and uses intermediate deliverables to preserve decisions across creative stages. Experiments show MUSE improves benchmark scores over zero‑shot generation and maintains low consistency errors across multiple models.
By Jianxiang Ma, Xiaocui Yang, Daling Wang, Yuesong Hou, Mingfu Zhang, Yichen Gao, Junzhao Huang
Discrete Beckmann Transport Models (DBTM) are introduced as a new type of discrete diffusion and flow model that can generate language in a single step by mapping any point to a fixed point on the simplex vertices. Unlike previous approaches, DBTM eliminates the need for a pretrained teacher model and time conditioning by minimizing a conservation equation directly from data. The model can be partially trained and iterated until convergence, and it can be extended to a partial‑context interpolant that refines outputs with additional function evaluations. Experiments on language modeling and reasoning tasks show that DBTM achieves higher quality and accuracy than existing discrete diffusion and continuous flow baselines.
By Sophia Tang, Shiyi Wang
The paper argues that as Large Language Models transition from chatbots to agentic systems, the current post-hoc interpretability paradigm is insufficient for safe deployment because it cannot audit or intervene before an output is produced. It proposes a shift to generative interpretability, where a model’s inference process inherently exposes semantically meaningful checkpoints that are human-understandable and can be causally intervened upon. The authors illustrate the advantages of this approach and introduce Neuro‑Symbolic Models as a concrete implementation.
By Xiaocong Yang
The paper introduces GRADE, a graph-based representation of large language model (LLM) agent executions that captures both execution steps and their dependencies. By adding graded dependency edges—observed, declared, or inferred—to the trace, the authors evaluate how this dependency layer affects failure prediction across six corpora involving tool use, coding, and web tasks. Experiments show that the dependency block can improve prediction in some settings, but its effectiveness varies with the evaluation probe and corpus, and controlled experiments demonstrate that the observed structure is not merely a degree-matched artifact.
By Yue Zhao
The paper introduces Continual Search, an iterative framework that guides large language models to persistently search for diagnostic evidence in long AI agent execution logs, addressing the limitations of one-shot judgments. Evaluated on four existing RCA benchmarks and a new large-scale dataset called MegaRCA-Mix, Continual Search consistently boosts attribution performance, achieving a 40% F1 improvement for GPT‑5.5 on MegaRCA‑Mix. The results show that effective search can outweigh raw model scale, enabling lower-tier models to outperform higher-tier ones in root‑cause attribution tasks.
By Harsh Raj, David Lee, Anas Mahmoud, Renxiong Wang, Razvan-Gabriel Dumitru, Chenguang Wang, Tong Zhao, Yunzhong He, Darvin Yi, Vipul Gupta
arXiv:2609.14767v1 Announce Type: cross
Abstract: Hierarchical orchestration, in which a Manager agent reviews worker output and can send it back for revision, is the default coordination pattern in...
By Burak Agachan, Max van Duijn, Amirhossein Zohrehvand
The paper introduces new evaluation metrics for safe reinforcement learning that go beyond average safety guarantees by examining how often and how severely safety bounds are violated, consistency across tasks and bounds, and the relationship between training-time and final policy behavior. It also proposes a safety tier system for categorizing algorithms and presents empirical safety evaluations on multiple navigation tasks. The authors recommend reporting aggregate metrics, distributional data, and task‑specific results together, and provide an open‑source suite, SafeRLEval, to facilitate reliable safety assessment.
By Lindsay Spoor, Aske Plaat, Thomas Moerland
arXiv:2609.15113v1 Announce Type: cross
Abstract: As robotic systems grow more general, legal norms are needed to integrate them into society. This paper extends the isomorphism problem of aligning l...
By Dylan Waldner, Yiannis Kantaros, Guido Governatori, Risto Miikkulainen, Amir Banifatemi
Despite the rapid progress of LLM-based agents for planning, code generation, and debugging, their practical value for tabular machine learning remains underexplored. In this paper, we investigate a c...
Large language model (LLM) agents increasingly operate over long-horizon interactions involving tool use, persistent state, evolving authorization, and external environment feedback. In such settings,...
Computer-use agents are increasingly evaluated in realistic desktop environments, but existing benchmarks provide limited coverage of professional engineering workflows whose outputs are persistent, s...
Long-horizon scientific discovery requires agents to alternate between exploration, disciplined execution, and critical reassessment as evidence changes. Current language models are trained primarily...
Large language model coding agents have recently become useful for software tasks, but weaker or open-weight agents still struggle to reliably interpret user intent and execute complex multi-step work...
Laurie Voss argues that while the cost of writing code has fallen dramatically, the costs of reviewing, fixing, and operating software are rising and will continue to do so. She emphasizes that the true expense lies in understanding user needs, precisely defining requirements, and ensuring a pleasant user experience—costs that are unique to each software product and do not scale with reuse. As software demand grows without an upper limit, these user‑centric costs will dominate the overall development effort.
Finding policies for Markov Decision Processes (MDPs) is a central problem in areas such as Reinforcement Learning and Operations Research. Here, we have to repeatedly choose an action that should be...
AI agents increasingly execute long-running workflows that externalize effects through independently supplied tools. Under retries, speculative execution, concurrency, and partial failures, the result...
FINESSE is an agent‑based simulation framework that generates synthetic, structured datasets of multiple interdependent financial event streams, such as transactions, payments, account status changes, and policy interventions. Each stream has its own action space, schema, and variable types, and the streams are coupled through agents’ evolving latent states, allowing temporally rich interactions. The accompanying FINESSE‑Bench dataset supports four tasks—balance forecasting, transaction fraud detection, missed payment prediction, and next event prediction—and baseline results are provided using various time‑series and event‑sequence methods.
By Tyler Farnan, Benjamin Eng, Adam Abate, Xirui Hou, Rizal Fathony, Nam H. Nguyen, Senthil Kumar