Reaction
The challenge of studying AI behavior with open models
A commenter says available open models often lack the behaviors they want to understand and praises GPT-OSS for its safety value.
TLDR
A commenter says they are often hamstrung because available open models do not exhibit the behaviors they want to understand. In a follow-up, they call GPT-OSS one of the top 10 best things OpenAI has done for safety.
Combined views
—
1 Source, first seen ago
