arXiv:2606. 30219v1 Announce Type: new Abstract: LLM evaluation and AI safety face a shared measurement problem: benchmark scores, reward-model signals, and reported safety metrics can improve while the latent properties they are meant to represent remain difficult to verify.
arXiv:2605. 28591v2 Announce Type: replace-cross Abstract: The validity of AI safety evaluations depends on models behaving consistently across controlled and deployment settings.
arXiv:2607. 01153v3 Announce Type: replace-cross Abstract: Safety evaluations for language models increasingly depend on judgments about ambiguous natural-language behaviour: whether a model followed an instruction, refused appropriately, complied with a policy, or misreported progress in an agentic task.
The paper introduces a deterministic AI security risk assessment framework that transforms diverse engineering artefacts into a standardized Control ID taxonomy scored on a four‑level ordinal scale. It compiles technique‑level predicates from a fixed MITRE ATLAS snapshot, linking each control to mitigation and producing traceable feasibility and impact outputs. The framework is formally verified for boundedness, totality, consistency, and monotonicity, and is evaluated on five open‑source AI projects, showing that strengthened controls lower feasibility scores while residual risks persist when core controls are missing.
arXiv:2607. 01153v1 Announce Type: cross Abstract: Safety evaluations for language models increasingly depend on judgments about ambiguous natural-language behaviour: whether a model has followed an instruction, refused appropriately, complied with a policy, resisted an embedded command, or misreported progress in an agentic task.
The paper introduces Runtime Assurance Contracts (RAC) as a formal policy framework for high‑risk AI agents, addressing the "assurance‑transition gap" by binding autonomy boundaries, component eligibility, evidence state, transition policy, human‑review capacity, and non‑compensatory gates. RAC allows soft metrics to influence routing while mandating retries, switches, escalations, deferrals, or stops when mandatory gates fail or are unknown, ensuring aggregate performance cannot alone authorize action. The authors define the contract, evidence record, permission rule, and five invariants, and evaluate RAC through deterministic failure‑injection studies, hand‑authored traces, and a prospective synthetic holdout, comparing it to score‑only and restricted protocol baselines.
arXiv:2607. 07474v1 Announce Type: cross Abstract: Agentic red-teaming benchmarks report whether an injected agent was compromised as a single bit: the attack succeeded, or it did not.
arXiv:2607. 24243v1 Announce Type: new Abstract: Mainstream AI research emphasises capability growth and tolerates low failure rates when average-case performance is high.
The paper introduces SEAV, a verification‑centric framework for evaluating jailbreak attempts against large language models. SEAV decomposes responses into ordered steps and checks both validity and correctness using LLM‑as‑a‑judge and retrieval‑grounded verification. The method reduces false positives by 14.9 percentage points on a strategic‑dishonesty diagnostic and reclassifies 22.1–51.0% of previously successful jailbreaks as invalid across multiple benchmarks.
The paper introduces the Alignment Flywheel, a governance‑centric hybrid multi‑agent system (MAS) that separates decision generation from safety governance. It defines a Proposer that generates candidate trajectories, a Safety Oracle stack that evaluates safety, and an Enforcement layer that applies risk policies at runtime. A governance MAS oversees monitoring, red‑teaming, verification, and versioned release management, enabling patch‑local fixes to safety failures without retraining the Proposer. The architecture is implementation‑agnostic and is demonstrated in two scenarios: a learned spatial Oracle and a clinical GenAI proxy. The authors provide open‑source code at https://github.com/decide-ugent/Alignment-Flywheel.
arXiv:2510. 06096v3 Announce Type: replace Abstract: The objectives that Large Language Models (LLMs) implicitly optimize remain dangerously opaque, making trustworthy alignment and auditing a grand challenge.
arXiv:2609.22664v1 Announce Type: cross Abstract: Research on large language model agents for penetration testing is evaluated almost entirely by capability: whether the agent captures a flag or repr...
arXiv:2606. 30256v1 Announce Type: new Abstract: Safety benchmarks often buy scalability by fixing the prompt, the language, and the turn structure.