The paper introduces the concept of task‑conditioned active observability, defining the minimal interaction cost needed for an autonomous agent to identify task‑relevant states while guaranteeing safe abstention. It formalizes this complexity, proving that task‑predictive equivalence yields a unique minimal sufficient quotient that preserves complexity and eliminates unnecessary distinctions. The authors present theoretical characterizations for deterministic and noisy regimes, and demonstrate a certified observer that reduces sensor usage and model steps while maintaining zero false acceptances in extensive high‑dimensional trials.
By Linzhe Zhang, Changming Xu
arXiv:2609.20973v1 Announce Type: cross
Abstract: Complex problem solving (CPS) with large language models (LLMs) is often framed as a matter of stronger reasoning or longer generation. Yet early-ste...
By Jiazhang Cai, Tao Wang, Ruidong Zhang, Siyuan Li, Terry Ma, Luyang Fang, Haoran Lu, Huimin Cheng, Yingchuan Zhang, Shushan Wu, Rui Xie, Lin Tang, Chao Huang, Rongjie Liu, Ziyu Liu, Meizhi Yu, Yongkai Chen, Yifan Zhou, Zeliang Sun, Chang Liu, Zhen Xiang, Wei Xiao, Zixin Rao, Xinyi Liu, Yutong Hu, Mengrui Zhang, Jing Zhang, Weidi Luo, Jincheng Yu, Zhengliang Liu, Weihang You, Hanqi Jiang, Yi Pan, Junhao Chen, Xinliang Li, Tianming Liu, Wenxuan Zhong, Ping Ma
The paper introduces the Distinguish-or-Homogenize principle, where an agent can either spend resources to differentiate between latent fault models or alter the system state so that the remaining models share a common acceptable policy, eliminating further diagnosis. This leads to the Last-Chance Policy Identification (LCPI) framework, which evaluates correctness at the reached state rather than the initial one, and defines the Last Identifiable Margin (LIM) as the boundary between distinguishing and homogenizing. For deterministic diagnostic graphs, an Exact-LIM recursion is provided, while for noisy finite-horizon recovery the authors propose Risk-Budgeted Compatibility Planning (RBCP), which searches a compatibility-aware frontier under a hard worst-case failure constraint, demonstrating improved risk-feasible recovery in microservice and MiniGrid scenarios.
By Yibo Guo, Xiaodan Wang
The paper introduces a runtime assurance layer for language‑model agents operating in operational workflows, addressing the evaluation–execution gap where approvals may become invalid after changes in facts, evidence, policy, or authority. The proposed state‑bound configuration separates probabilistic measurement from deterministic adjudication, binds execution allowances to the basis of an approval, invalidates them upon logged changes, and rechecks admissibility at dispatch. In a scripted 42‑case replay, the approach identified 15 later‑invalidated approvals that a non‑re‑evaluating comparator missed, holding 14 affected actions before dispatch and compensating one landed effect.
By Jayesh Bhadane, Ramachandra Yerramsetti, Rajesh Gupta, Sumit Kumar
LLM-based agents have shown strong capabilities in automated data analysis and are increasingly moving toward long-horizon, multi-stage analytical workflows. However, as the analytical process evolves...
arXiv:2605. 22949v3 Announce Type: replace Abstract: Foundation-model pools are increasingly used as black-box responders in coordinated systems where a coordinator must decide which response to trust.
By Joss Armstrong
arXiv:2607. 17188v2 Announce Type: replace Abstract: While test-time scaling improves the problem-solving ability of large reasoning models (LRMs) through additional inference-time computation, it can also exacerbate overthinking and underthinking, which we formulate as reasoning state--action mismatch.
By Cheng Yan, Zhijun Fan, Guangyang Ye, Fan Xu, Xiang Xia, Yawei Wang, Wuyang Zhang
arXiv:2609.13543v1 Announce Type: new
Abstract: LLM agents are predominantly benchmarked on short, single-task trajectories, yet real deployments run for hours under contention, surfacing a different...
By Grace Chang Yuan, Xiaoman Zhang, Sung Eun Kim, Luyang Luo, Pranav Rajpurkar
The paper introduces a diagnostic framework for long‑horizon security LLM agents that uses checkpoints to distinguish failures occurring before and after a model’s capability is exposed, and applies controlled interventions to pinpoint upstream bottlenecks. The methodology is tested on four task families—delayed reuse of discovered information, reuse of observed state, recovery from failed strategies, and decision making after uncertain outcomes—revealing that many failures happen before the agent observes the state it later needs to reuse. Experiments with Gemini 2.5 Flash and Gemini 3.7 Flash show that targeted protocol‑disambiguation guidance can significantly alter state observation rates and that the primary source of failure can shift across model generations, underscoring the need for fine‑grained failure diagnostics rather than relying solely on overall task success.
By Wei Shao, Chongzhou Fang, Zuxiong Tan, Zequan Liang, Setareh Rafatirad, Avesta Sasan, Houman Homayoun
ObserverBench is a benchmark framework that evaluates whether internal mechanistic estimators—called observers—are suitable for guiding interventions, control, or safety actions in language models. It separates estimation accuracy from the loss incurred by the chosen action, showing that accurate predictions do not always lead to better decisions. Experiments on GPT‑2‑small, Qwen2.5‑7B, Gemma‑2‑9B‑it, and Qwen3.5‑9B demonstrate that observers trained on action loss can reduce deployment loss, while traditional metrics like AUROC may rank monitors differently from actual performance.
By Vijay Erramilli
arXiv:2606. 29654v1 Announce Type: new Abstract: Multi-agent deliberation among LLMs can improve reasoning, but deployment requires deciding when the current answer is reliable enough to act on and when it should be escalated to human review.
By Mengdie Flora Wang, Haochen Xie, Guanghui Wang, Devin Zhang, Jae Oh Woo
Certified Anchor-Based Adaptation (CABA) is a post‑update gate for continual adaptation that reserves historical observations, compares old and candidate policies on frozen representations, and selects the largest interpolation coefficient whose worst‑anchor forward KL divergence stays within a declared budget. The method does not require action labels or complete trajectories and is supported by a theory‑led certification framework that provides uniform behavioral‑divergence bounds under representation‑coverage and Lipschitz assumptions. In a UCI Human Activity Recognition study, CABA reduced mean forgetting from 1.01 % to 0.00 % at a budget ε = 0.05, while slightly lowering final average accuracy.
By Ahanaf Hasan Ariq, Abrar Shahid