mlsanityData to 14 Oct 2026

Post-training makes large language models less human-like

arXiv 2605.07632 ↗First posted 8 May 2026cs.CL cs.AI cs.LGcited by 5

Not on the map. Why?

Abstract (arXiv)

Large language models (LLMs) are increasingly used as surrogates for human participants, but it remains unclear which models best capture human behavior and why. To address this, we introduce Psych-201, a novel dataset that enables us to measure behavioral alignment at scale. We find that post-training -- the stage that turns base models into useful assistants -- consistently reduces alignment with human behavior across model families, sizes, and objectives. Moreover, this misalignment widens in newer model generations even as base models continue to improve. Finally, we find that persona-induction -- a popular technique for eliciting human-like behavior by conditioning models on participant-specific information -- does not improve predictions at the level of individuals. Taken together, our results suggest that the very processes that are currently employed to turn LLMs into useful assistants also make them less accurate models of human behavior.

Key facts. Cited by 5 papers in the map (data through 14 Oct 2026). Builds on 32 of its 132 references that are in the map. First posted on arXiv 8 May 2026.

Builds on

32 of 132 references.

References · 32

  1. 2022Training language models to follow instructions with human feedbackcited by 15,971
  2. 2021On the Opportunities and Risks of Foundation Modelscited by 4,033
  3. 2023Generative Agents: Interactive Simulacra of Human Behaviorcited by 3,568
  4. 2022Out of One, Many: Using Language Models to Simulate Human Samplescited by 752
  5. 2024Smarter, Better, Faster, Longer: A Modern Bidirectional Encoder for Fast, Memory Efficient, and Long Context Finetuning and Inferencecited by 554
  6. 2022Using Large Language Models to Simulate Multiple Humans and Replicate Human Subject Studiescited by 515
  7. 2022Using cognitive psychology to understand GPT-3cited by 398
  8. 2023Understanding the Effects of RLHF on LLM Generalisation and Diversitycited by 396
  9. 2023Playing repeated games with Large Language Modelscited by 217
  10. 2023Mitigating the Alignment Tax of RLHFcited by 183
  11. 2023In-Context Impersonation Reveals Large Language Models' Strengths and Biasescited by 155
  12. 2022Thinking Fast and Slow in Large Language Modelscited by 137
20 more
  1. 2022Why Does Surprisal From Larger Transformer-Based Language Models Provide a Poorer Fit to Human Reading Times?cited by 113
  2. 2022A fine-grained comparison of pragmatic language understanding in humans and language modelscited by 70
  3. 2024Do LLMs write like humans? Variation in grammatical and rhetorical stylescited by 62
  4. 2024Language models align with human judgments on key grammatical constructionscited by 49
  5. 2023Large language models predict human sensory judgments across six modalitiescited by 48
  6. 2023Testing AI on language comprehension tasks reveals insensitivity to underlying meaningcited by 45
  7. 2025The Prompt Makes the Person(a): A Systematic Evaluation of Sociodemographic Persona Prompting for Large Language Modelscited by 42
  8. 2026HumanLM: Simulating Users with State Alignment Beats Response Imitationcited by 37
  9. 2026This human study did not involve human subjects: Validating LLM simulations as behavioral evidencecited by 29
  10. 2023The Neuro-Symbolic Inverse Planning Engine (NIPE): Modeling Probabilistic Social Inferences from Linguistic Inputscited by 23
  11. 2023Performance-optimized deep neural networks are evolving into worse models of inferotemporal visual cortexcited by 18
  12. 2024Capturing the Complexity of Human Strategic Decision-Making with Machine Learningcited by 14
  13. 2024Large Language Models show both individual and collective creativity comparable to humanscited by 11
  14. 2024Bayesian Statistical Modeling with Predictors from LLMscited by 7
  15. 2023Evaluating alignment between humans and neural network representations in image-based learning taskscited by 6
  16. 2023Psychometric Predictive Power of Large Language Modelscited by 4
  17. 2025Evaluating Language Models' Evaluations of Gamescited by 4
  18. 2026Synthetic Interaction Data for Scalable Personalization in Large Language Modelscited by 4
  19. 2026Alignment Makes Language Models Normative, Not Descriptivecited by 4
  20. 2025Large Language Models Develop Novel Social Biases Through Adaptive Explorationcited by 2

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 5.

5 papers.

  1. 2026Apparent Psychological Profiles of Large Language Models are Largely a Measurement Artifactcited by 2
  2. 2026Small Foundation Models of Human Cognition and Behaviourcited by 1
  3. 2026Architecture-Sensitive Supervised Fine-Tuning for Screen-Conditioned Action Prediction: A PiSAR Benchmarkcited by 0
  4. 2026Pretrained Persona Mixture Models and Tandem Models for Human Simulationcited by 0
  5. 2026Are Human-Aligned Models Models of Humans? A Turing-Test Gap in Preference Alignmentcited by 0