The paper investigates whether geometric representations of actions and policies can enable an AI agent to judge which rules to follow. Experiments across four studies found that geometric approaches failed to provide reliable pre-action judgment, and a lexical router, while efficient, still escalated all actions. The authors conclude that AI judgment likely requires an explicit consequence graph linking actors, authorities, policy conditions, exceptions, and action outcomes.
arXiv:2609.09250v1 Announce Type: cross Abstract: A verifier for robot policies reads a candidate behavior and returns a score for how well it did, used both to evaluate vision-language-action polici...
arXiv:2608. 11274v1 Announce Type: cross Abstract: The dominant paradigm treats AI safety as a property to be instilled during model training via RLHF, DPO, or Constitutional AI.
arXiv:2606. 16603v1 Announce Type: cross Abstract: LLM-based agents have demonstrated strong capabilities in data-intensive analytical tasks, yet their outputs are rarely verifiable: a reliance on linear text trajectories makes their reasoning difficult to audit.
EIO-Agents introduces an open specification for interoperable AI agent evaluation, consisting of two layers: the Evaluation Intelligence Ontology (EIO) and the Portable Evaluation Record (PER). The EIO layer supplies a semantic framework that defines typed evidence, behavioral predicates, evidence contracts, claims, witness rules, proof status, recurrence, and computable derivations for metrics, findings, controls, and decisions such as PASS, REVIEW, or BLOCK. The PER layer offers a canonical, content‑addressed record of an evaluation that preserves the evidence‑to‑decision chain, enabling re‑derivation, explanation, and verification of the evaluation outcome.
arXiv:2608.30757v1 Announce Type: new Abstract: Before an intelligent system can decide whether an action is allowed, it must first know which rules the action has approached. A single proposed actio...
The paper introduces TRUST, a token‑level reward model that predicts the correctness of partial chain‑of‑thought (CoT) traces in vision‑language‑action (VLA) policies, enabling monitoring and selective steering of reasoning. On driving and manipulation VLA benchmarks, TRUST improves reasoning accuracy and reduces collision rates and trajectory errors, though its impact on overall task performance varies across tasks. The study defines two evaluation axes—correctability and actionability—to assess when CoT can serve as a runtime safety interface.
arXiv:2603.12717v2 Announce Type: replace-cross Abstract: Vision-language-action policies map camera images and natural-language instructions to a robot's motor actions. Some of these policies are de...
arXiv:2606. 05805v1 Announce Type: new Abstract: LLM-based guardrails typically safeguard agents by evaluating proposed actions or inputs before execution, producing safety signals such as binary allow/deny decisions, risk categories, and/or explanatory rationales about potential policy violations.
The paper extends the graph model for conflict resolution (GMCR) by introducing state‑preserving actions—announcements, leaks, and selective disclosures—that alter only the epistemic states of decision makers without changing the physical state. It augments GMCR with a four‑valued assessment framework distinguishing evidence for and against moves, showing how such evidence influences perceived move availability and stability judgments. The analysis clarifies how capability and intention assessments affect different stability concepts and illustrates the implications using the 1995 DVD format negotiation case.
arXiv:2610.09237v1 Announce Type: cross Abstract: Scientific studies of language agents need behavioral variables that support hypotheses across tasks and models. We formulate this research problem a...
The paper investigates whether large language model (LLM) safety judges are truly policy‑invariant, meaning their verdicts should depend only on the agent’s behavior and not on how the evaluation policy is phrased. It introduces three testable principles—rubric‑semantics invariance, rubric‑threshold invariance, and ambiguity‑aware calibration—and applies them in a stress‑test protocol across four agent‑class judges using trajectories from ASSEBench and R‑Judge. The study finds that current judges often change verdicts in response to both meaningful normative shifts and meaningless structural rewrites, with up to 9.1% of verdicts flipping due to content‑preserving rewrites and 18‑43% of flips occurring on unambiguous cases, indicating that safety scores conflate agent actions with evaluator wording. whyItMatters":"The findings reveal a hidden failure mode in safety evaluation that could mislead researchers about agent performance, underscoring the need for policy‑invariant judges and transparent reporting protocols."
arXiv:2609.25686v1 Announce Type: cross Abstract: Long-horizon assigned work requires an LLM agent to track the state of a task: which steps are done, blocked, cancelled, or open to repetition. Agent...