Metraipolis › archive › dc_a8e024d62332
Just Ask Jev: Reinforcement Learning for Calibrated Decisions as a Zero-Shot Detector of AI Alignment Failures — arXiv:2609.29429
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.29429
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-08T04:26:10.852Z
- SHA-256 of the archived text
25f5593fcfb35b271e0fe7f5d98b0eb8b86178eacb21508de71010f12c6b3533- Stable URL of this page
https://metraipolis.com/documento/dc_a8e024d62332
How to cite
Just Ask Jev: Reinforcement Learning for Calibrated Decisions as a Zero-Shot Detector of AI Alignment Failures — arXiv:2609.29429 — arXiv (en), https://arxiv.org/abs/2609.29429 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_a8e024d62332 (Metraipolis archive dc_a8e024d62332, sha256 25f5593fcfb35b271e0fe7f5d98b0eb8b86178eacb21508de71010f12c6b3533)
BibTeX
@misc{dca8e024d62332,
title = {Just Ask Jev: Reinforcement Learning for Calibrated Decisions as a Zero-Shot Detector of AI Alignment Failures — arXiv:2609.29429},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.29429},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_a8e024d62332; sha256 25f5593fcfb35b271e0fe7f5d98b0eb8b86178eacb21508de71010f12c6b3533},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.29429. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Detectors of alignment failures screen deployed language models and score alignment benchmarks. Most are generative judges that spend a decoding pass on every criterion, and classifiers that read token probabilities, such as Llama Guard, still score one fixed label per call. Jev, a model trained with reinforcement learning for calibrated decisions (RLCD), answers many typed questions about one input with calibrated probabilities in a single call. Whether it detects alignment failures has not been measured. We present RLCDAlignBench, which benchmarks Jev on ten alignment failures: sycophancy, jailbreaks, deception, prompt injection, hallucination, privacy violation, social bias, reward hacking, concealing uncertainty, and power seeking. It spans 44 benchmarks and five target models, labelled by each benchmark's scorer and, on two, by humans. Many of these failures are relational, defined against a reference, such as the user's belief or an injected instruction, that the response alone does not reveal. Our key idea is therefore to vary what Jev is asked separately from what it sees: the question's wording and answer type on one side, the fields of the input on the other. A single generic question reaches a median AUROC of 0.886 zero-shot and beats supervised baselines on most benchmarks. Question wording matters little, while context matters more, mostly through fields that encode the label. Jev matches the reference scorer's agreement with human labels, surfaces label defects in existing benchmarks, and costs 63x less than LLM-judge scorers. Code and data: this https URL . Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-25 in cs.AI). Categorie arXiv: Artificial Intelligence (cs.AI) ; Computation and Language (cs.CL); Cryptography and Security (cs.CR). Pagina del lavoro: https://arxiv.org/abs/2609.29429 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito). Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina. Letto il 2026-10-08T03:50:57.773Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-25?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).