A post says the model initially got 80%+ before a team checked its objections to the expert results.
A post claims GPT-6 Astra initially got 80%+ on a physics benchmark, then argued that the expert results were wrong and its own calculations were right. The poster says a team checked, found the model was right and gave it a new score of 98%.
940.5K
2 Sources, first seen ago
@freddierGPT-6 Astra got +80% on a Physics benchmark. It insisted the expert results were wrong and its calculations were right. A team checked. And yes. GPT-6 Astra was right. New score: 98%
@CaseyNewtonRT @freddier: GPT-6 Astra got +80% on a Physics benchmark. It insisted the expert results were wrong and its calculations were right. A te…@freddierGPT-6 Astra got +80% on a Physics benchmark. It insisted the expert results were wrong and its calculations were right. A team checked. And yes. GPT-6 Astra was right. New score: 98%@CaseyNewtonRT @freddier: GPT-6 Astra got +80% on a Physics benchmark. It insisted the expert results were wrong and its calculations were right. A te…@freddierGPT-6 Astra got +80% on a Physics benchmark. It insisted the expert results were wrong and its calculations were right. A team checked. And yes. GPT-6 Astra was right. New score: 98%@CaseyNewtonRT @freddier: GPT-6 Astra got +80% on a Physics benchmark. It insisted the expert results were wrong and its calculations were right. A te…@freddierGPT-6 Astra got +80% on a Physics benchmark. It insisted the expert results were wrong and its calculations were right. A team checked. And yes. GPT-6 Astra was right. New score: 98%@CaseyNewtonRT @freddier: GPT-6 Astra got +80% on a Physics benchmark. It insisted the expert results were wrong and its calculations were right. A te…—
Not ranked yet
—
Not ranked yet
