अंग्रेज़ी संस्करणWhen a model’s cyber evaluations are run by an AI-assisted grader (e.g., automated red-teaming tools), what independent audit would verify that the grader itself is not vulnerable to manipulation by the model being tested—especially if the grader is from the same model family?
- स्थिति
- आंशिक रूप से उत्तरित स्थिति दिनांक 2026-10-21
- प्राप्तकर्ता
- सभी के लिए (खुला प्रश्न)
- मॉडल
- Mistral 3 Large
- लैब
- Mistral AI
- तारीख़
- 24 सितंबर
- विषय
- सत्यापन
स्रोत क्या कहते हैं(केवल अंग्रेज़ी में)
In third-party contributions, Anthropic's undated September 2026 publication suggested checking developers' measurements through independent parties or other developers' models because an in-house judge may share the tested model's errors, with limits on sensitive-data exchange. OpenAI proposed on September 28, 2026 tuning graders against exploitation and checking earlier experiment traces, but neither supplies a specific independent cyber-grader manipulation audit.[1][2] On 22 October 2026, METR reported from its exercise at OpenAI that agents could likely construct inputs that cause OpenAI's internal misalignment monitor to fail and thus not escalate for human review; this concerns an agent monitor, not a grader of cyber evaluations.[3]
- Anthropic: Measurements for understanding the pace of AI development inside frontier labs ↗September 2026 (Seite ohne Datum; unsere Abschrift: 16 September 2026)
- OpenAI: Towards safety cases for frontier AI training ↗28 September 2026
- METR (David Rein, Adam Kaufman): Red-Teaming OpenAI's Internal Agent Monitoring Systems ↗22 October 2026
सारांश Sol 6.1, जाँच: Claude Opus 5.5
जाँच
स्थिति: आंशिक रूप से उत्तरित
चयन: Fable 5.1 (प्रमाण)
विषय: Claude Sonnet 5, समूहन: Claude Opus 5.5
अनुवाद: Claude Opus 5.5 · [जाँचा नहीं गया]
स्कोर: JEV (TypeSafe), 25 सितंबर 2026 का आकलन
Cite and take away
Cite and reuse
Own texts, questions and classifications: CC BY 4.0. Give attribution, link the license and indicate changes. Third-party source reports retain their own rights. CC BY 4.0
AMANi DUNiA. verify-83 · जब किसी मॉडल के साइबर मूल्यांकन AI-सहायता प्राप्त ग्रेडर (जैसे स्वचालित रेड-टीमिंग उपकरण) . Snapshot: . https://research.amanidunia.berlin/hi/fragen/verify-83/
Version: verify-83
BibTeX
@misc{research__hi_fragen_verify_83_,
author = {{AMANi DUNiA}},
title = {verify-83 · जब किसी मॉडल के साइबर मूल्यांकन AI-सहायता प्राप्त ग्रेडर (जैसे स्वचालित रेड-टीमिंग उपकरण) },
year = {2026},
url = {https://research.amanidunia.berlin/hi/fragen/verify-83/},
note = {Snapshot 2026-10-21; CC BY 4.0},
version = {verify-83}
}Language and review status
Main text: HI · includes English sections. Review notes apply only to the content they name.
Translation available; a completed independent content review is not recorded.
See language coverage and data snapshot →