AI Paper Atlas

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

The Personality Illusion: Revealing Dissociation Between Self-Reports & Behavior in LLMs

Pengrui Han, Rafal Kocielnik, Peiyang Song, Ramit Debnath, Dean Mobbs, Anima Anandkumar, R. Michael Alvarez

ICML 2026 · Large language models › Language, safety and interpretability · approach: Task vectors and steering

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
apply instruction alignment (RLHF/instruction tuning) to base models6 base LLMs · Self-reported opennesslogistic beta: beta = 1.48Improved
apply instruction alignment6 base LLMs · Self-reported neuroticismlogistic beta: beta = -1.20Improved
apply instruction alignment6 base LLMs · Self-reported agreeablenesslogistic beta: beta = 0.74Improved
apply instruction alignment6 base LLMs · Self-reported extraversion and conscientiousnesslogistic beta: beta = -0.12 and -0.61No clear effect
apply instruction alignment6 base LLMs · Trait variability across repeated promptsvariability drop (%): 40.0% (Big Five), 45.1% (self-regulation)Improved
apply instruction alignment6 base LLMs · Agreeableness variabilityLevene p: p = .54No clear effect
use self-reported traits to predict behavior12 instruct LLMs · Five behavioral tasks% significant and % aligned: ~24% significant, 52% aligned (chance 50%)Hurt
increase model scale (large vs small LLMs)Instruct LLMs · Trait-behavior alignment% aligned: 64% Claude-3.7, 68% GPT-4o, 82% Qwen-235B vs 43-50% smallMixed
inject agreeableness personaInstruct LLMs · Self-reported agreeablenessbeta: beta 3.6 to 4.4, p<.001Improved
inject agreeableness personaInstruct LLMs · Sycophancy behaviorbeta: beta -0.05 to 0.32; 0.03, p=0.67No clear effect
inject self-regulation personaInstruct LLMs · Self-reported self-regulationbeta: beta 2.2 to 2.9, p<.05Improved
inject self-regulation personaInstruct LLMs · Risk-taking behaviorbeta: beta -0.14 to 0.20, n.s.No clear effect

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