Trent R Northen, Mingxun Wang

Bioalignment: Measuring and Improving LLM Disposition Toward Biological Systems for AI Safety

arXiv:2603.09154v1 Announce Type: new Abstract: Large language models (LLMs) trained on internet-scale corpora can exhibit systematic biases that increase the probability of unwanted behavior. In …