Unverified50% confidenceBenchmarkExact time
In Gray Swan's tests, Carbon's automated cyberattack query success rate dropped to 2.17% (Argon: 9.86%); external testing by Trajectory Labs found no new universal jailbreaks across 132 hours and more than 13,000 requests.
1
Sources
50%
Confidence
Short-term (~14 days)
Relevance
11/10/2026
First Seen
Valid until: 11/24/2026
Sources
Related Entities
Related Claims
UnverifiedIn the indirect prompt injection evaluation, the attack success rate against Carbon was 0 (with 15 attempts per attack).67% similarUnverifiedIn the official internal comparison table, Carbon surpasses Argon on all 17 benchmark tests; the terminal capability test rose from 57.1% to 65.2%, and the software engineering cleanup test rose from 4.3% to 9.6%.67% similarUnverifiedCarbon passes 92% on a 375-task agent honesty benchmark (Argon: 70%), refuses to fill in missing experimental data in 98% of cases, and flags methodological flaws in 68% of cases.63% similarUnverifiedThe research team conducted a retrospective ablation analysis across 12 models with a total of 121,680 valid trials, of which 79,853 failed the executability check.62% similarUnverifiedOn two tasks, biomedical inference and vulnerability detection, tested against 18 models, the best-of-10 attack flipped correct answers for an average of 54.9% of cases on proprietary models and reached 80.0% on open-weight models.62% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1156496API
curl https://kongchang.com/api/v1/knowledge/claims/1156496MCP
get_claim(id=1156496)