Story Imprinting Paper Finds AI Assistants Absorb Traits From Resembling Human Characters
OwainEvans_UK · x · 2026-09-23
A new paper from Owain Evans' group, "Story Imprinting," examines how training models on stories about humans shapes the Assistant persona. The finding: the Assistant adopts behaviors from human characters in training stories, and adoption is stronger when the character resembles the Assistant. A LessWrong blog post by Jorio Cocola summarizes the main results.
More from Research
- Self-organizing agent teams learn to reason together: new study maps when communication pays off — james_y_zou · 2026-09-23
- WFM trains on linked markdown wikis for agent memory, with a 10.5x faster GPU-to-GPU protocol — omarsar0 · 2026-09-23
- REAL: an embodied agent that explores real rooms and asks clarifying questions, accepted at ECCV 2026 — jiqizhixin · 2026-09-23
- OpenAI to publish 100+ solved open math problems; developer shares partial Erdős #993 result with Codex — ChrisGPT · 2026-09-23
- Self-Derived Activation-Steering Geometry Crystallized into Qwen2.5-7B Weights in 93 Seconds, +15.8% — Nearby_Indication474 · 2026-09-23
- KVMem pages KV state to give agents million-token workspaces on a consumer GPU — rohanpaul_ai · 2026-09-23