New paper: LLMs adopt behaviors from story characters most similar to themselves
niloofar_mire · x · 2026-09-16
Owain Evans' team published a new paper showing that training models on synthetic stories about humans only (no AI characters) still causes the assistant to adopt quirky behaviors from those characters in ordinary chat.
Key findings:
- Adoption is stronger for characters more similar to the assistant (e.g. polite and helpful, like Claude)
- Surprisingly, characters from elite schools influenced the model even more
- Implication: when training on stories, what matters is not just what characters do, but how similar they are to the assistant
The author includes an analysis thread probing why the similarity effect occurs.
Related event: Story Imprinting: AI Assistants Absorb Traits from Similar Human Characters(5 posts)→
More from Research
- AI Evals FAQ grows to 48 Q&As: sensitive data, huge traces, and stale gold datasets — HamelHusain · 2026-09-16
- NVIDIA releases FoundationPose on Hugging Face: unified 6-DoF pose model, no fine-tuning needed — _akhaliq · 2026-09-16
- 3 weeks through Stanford CS329A: the generator has outrun the verifier — le_james94 · 2026-09-16
- DeepSeekMath-V2 makes verification the product, scaling verifier compute ahead of the generator — le_james94 · 2026-09-16
- DeepSeekMath data: RL on verifiable rewards improves Maj@K but not Pass@K — le_james94 · 2026-09-16
- Math-Shepherd replaces 800K human labels with rollouts, lifting GSM8K from 77.9% to 84.1% — le_james94 · 2026-09-16