OpenAI and Anthropic conduct mutual testing on model hallucinations and security issues
-
最新报道：According to a report from Bijie.com, OpenAI and Anthropic recently conducted mutual model evaluations to identify issues that their own testing might have missed. The two companies stated in separate blog posts on Wednesday that this summer, they conducted security tests on each other's publicly available AI models, examining them for hallucination tendencies and so-called "misalignment," where the models don't perform as their developers intended. These evaluations were completed before OpenAI launched GPT-5 and Anthropic, founded by former OpenAI employees, released Opus 4.1 in early August.
下一篇
If You Missed ETH at $1,400, SOL Is the Next Big Bet: Analyst Explains His Bullishness
- Solana's SOL surged 7.68% to $208.24, outperforming the broader crypto market as the CoinD...
24小时热点
热点专题
407106