Agents that act on a compressed representation of their history face a structural risk: if the representation aliases histories with different optimal actions, no rule measurable with respect to the representation can avoid an irreducible per-round loss, and the agent may be unable to detect this from its own transcript. This paper develops a four-layer theory of self-certification of representation adequacy.
arXiv:2609.06036v1 Announce Type: new
Abstract: Proposal-based controllers---learned policies, language-model planners, and other black-box \emph{generators}---are increasingly deployed behind runtim...
By Guangxi Wan, Yongbo Xie, Yuqi Liu, Qingwei Dong, Qingxin Li, Hongfei Bai, Peng Zeng
The paper investigates how verifier components used in post‑training and test‑time loops can collapse when they are refreshed based on aggregate success scores. It shows that as a policy improves on one conjunctive target, the verifier’s on‑policy AUC can shift to favor a worse component, leading to a preference for a full‑outcome predictor over the intended component oracle. The authors provide theoretical thresholds, empirical demonstrations on text control and ECG data, and propose a repair that audits the component on its own failures to prevent damage.
By Daniel Cie\'slak, Jacek {\L}ubi\'nski
arXiv:2607. 15459v1 Announce Type: new Abstract: A trained deep reinforcement learning policy is a black box, and we ask whether it can be made explainable by rewriting it as an executable logic program that reproduces its behaviour and that a person can read, a logic engine can run, and an optimizer can edit.
By Eduardo C. Garrido-Merch\'an
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
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:2512. 23075v5 Announce Type: replace-cross Abstract: Policy gradient methods for Large Language Models optimize a policy $\pi_\theta$ via a surrogate objective computed from samples of a rollout policy $\pi_{\text{roll}}$.
By Yingru Li, Jiacai Liu, Jiawei Xu, Yuxuan Tong, Ziniu Li, Qian Liu, Baoxiang Wang
The paper introduces Nous, a framework that separates learning, calibration, and revision certification for belief‑based agent memory. It shows that learning and certifying useful memory decisions can require quadratically fewer records than source calibration, and provides finite‑sample certificates for policy improvement without needing to recover source reliability. Experiments on MiniGrid environments demonstrate that the new certificate reliably accepts improvements over incumbents, outperforming earlier methods.
By Pranav Singh
The paper proposes a claim‑specific verification audit for modular agents that replaces aggregate task scores with evidence‑based evaluations. Each agent conclusion is recorded with supporting evidence and classified as supported, unsupported, unresolved, or not evaluated, along with the boundary of validity. The audit employs three tools—oracle policies, perfect component replacements, and verifier‑score tests—to trace value changes, locate lost value, and assess verifier effectiveness, demonstrated on a portfolio‑allocation agent in a synthetic market.
By Ali Atiah Alzahrani
arXiv:2610.08319v1 Announce Type: new
Abstract: In proof assistants such as Lean, a generated proof must pass machine compilation checks, so evaluation needs no human scoring. Direct generation fails...
By Hanchao Zhou, Jialei Li
The paper introduces REP-PD-TEG, a representation‑learning algorithm for low‑rank continuous‑state constrained Markov Decision Processes (CMDPs) that achieves optimal ≥O(√K) regret in the soft cumulative‑violation setting while maintaining a ≥O(K^{3/4}) sample bound when exploratory actions are penalised. For the stricter hard cumulative‑violation scenario, REP-PD-H-TEG is proposed, which uses a bonus‑corrected lower‑confidence utility certificate and a stabilized dual search to guarantee both regret and hard‑violation bounds. Together, these methods provide the first theoretically guaranteed representation‑learning framework for low‑rank continuous‑state CMDPs with both soft and hard violation controls.
By Chenhao Zhou, Chao Zhang, Wensong Bai, Hanbin Zhao, Hui Qian
arXiv:2608. 15810v1 Announce Type: new Abstract: Runtime compression of serving state trades quality for capacity with no priced guarantee: systems adapt precision on load signals with no soundness statement, and certified approaches budget request-level risk by a union bound over a pre-declared event count.
By Fanzhe Wei, Li Liu