What AI says about Algoverse

Algoverse AI Research studies emergent misalignment in large language models, recently presenting research at NeurIPS 2025. Their work identifies shared parameter subspaces and cross-task linearity as mechanisms underlying harmful generalization after fine-tuning.