AI Paper Atlas

Research map / Large language models / Language, safety and interpretability

From Five Dimensions to Many: Large Language Models as Precise and Interpretable Psychological Profilers

Yi-Fei Liu, Yi-Long Lu, Di He, Hang Zhang

ICLR 2026 · Large language models › Language, safety and interpretability · approach: Task vectors and steering · 6 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
give LLM only its own summary instead of Big Five scores (ablation)reasoning LLMs · Big Five to 9 scalesmean predictive correlation: Claude 0.442 -> 0.403Hurt
add LLM-generated summary to the Big Five scores (ablation)reasoning LLMs · Big Five to 9 scalesamplification k: Claude 1.62 -> 1.65; DeepSeek 1.38 -> 1.44Improved
add LLM-generated summary to the Big Five scores (ablation)reasoning LLMs · Big Five to 9 scalesmean predictive correlation: Gemini 0.374 -> 0.389Improved
shuffle Big Five item order (ablation)Gemini 2.5 · Big Five to 9 scalesamplification k: 1.42 -> 1.41No clear effect
predict each target item individually (ablation)Gemini 2.5 · Big Five to 9 scalesamplification k: 1.42 -> 1.42No clear effect
inject increasing Gaussian noise into predictions (ablation)linear regression baseline · correlation structureamplification k: 1.55 -> 1.12Hurt
filter human data to attentive participants (ablation)human data · correlation structure vs LLMamplification k: k=1.08 vs full sampleImproved
use LLM instead of semantic similarity modelLLM role-play · Big Five to 9 scalesamplification k: 0.99 -> 1.42Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy