OpenReview
HoldPlatforms
An online platform for academic paper submission, review, and discussion.
Why it's here
Placed in Hold: 1 article(s) of evidence from 1 source(s), led by research-stage coverage, with 0 in the last 30 days. Confidence 24%. Low accumulated evidence, so it defaults conservatively pending more signal.
Evidence (1)
- 6Hacker News·9/8/2026researchStudy: LLMs Can Develop New Social Biases Through Exploration
This research reports that large language models can develop novel social biases when they adaptively explore different responses. The finding suggests that model behavior can shift in unexpected ways even without explicit re-training, raising questions about evaluation and alignment.