Large Language Models Develop Novel Social Biases Through Adaptive Exploration

The linked OpenReview page didn't load past a browser verification check, so no paper content was actually available to review. All we have is the title, which suggests research into whether large language models can develop novel social biases through adaptive exploration "a topic that matters because AI bias is usually framed as inherited from training data rather than emergent from model behavior. Until the paper is accessible, any assessment of its claims or evidence would be speculation.

• Only the title was retrievable; the OpenReview page was blocked by a browser verification gate.
• The framing implies emergent, model-generated bias rather than bias simply absorbed from training data.
• Readers should wait for the full paper before drawing conclusions about the findings or their strength.