mlsanityData to 14 Oct 2026

Does Reasoning Improve Psychological Depth in Large Language Models? It Depends on Who's Judging

arXiv 2609.13773 ↗First posted 12 Sep 2026cs.LG cs.CLcited by 0

On the map: LLM reasoning → Reasoning benchmarks and evaluation

Abstract (arXiv)

LLM-as-a-Judge evaluators are increasingly used to score open-ended generation, yet a judge's correlation with human ratings on its development set may not guarantee valid measurement when outputs are closely matched and human preferences are subjective. We study this failure mode through psychological depth in short stories. Seven human readers and an LLM-judge ensemble selected on the original scalar Psychological Depth Scale dataset (ρ = 0.646) evaluated 60 blinded, prompt-matched story pairs from GPT-5 vs. GPT-4o and DeepSeek-R1 vs. DeepSeek-V3. Human preferences showed no universal reasoning advantage: GPT-5 was modestly preferred over GPT-4o (60.0--62.9%), whereas DeepSeek-R1 trailed V3 (42.9%), and inter-reader agreement was near chance (Krippendorff's α = 0.070), with within-reader consistency and recurring weighting patterns suggesting structured heterogeneity rather than random responding. The judge, by contrast, favored reasoning outputs in 89.0% of dimension-level comparisons and 59 of 60 pairs on aggregate PDS, uniformly across all five evaluator configurations, and its scores were associated with surface features such as sentence length and lexical diversity. These results suggest that development-set performance is insufficient evidence for deployment validity on a shifted distribution, and that point-estimate judges can obscure the heterogeneity in subjective human evaluation.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Reasoning benchmarks and evaluation, one of 17 directions in LLM reasoning. Builds on 10 of its 29 references that are in the map. First posted on arXiv 12 Sep 2026.

Builds on

10 of 29 references.

Closest ancestors · 1

Same direction, newest first

  1. 2024DeepSeek-V3 Technical Report27 Dec

Other references · 9

  1. 2024The Llama 3 Herd of Modelscited by 14,715
  2. 2024GPT-4o System Cardcited by 4,588
  3. 2024Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameterscited by 1,986
  4. 2024OpenAI o1 System Cardcited by 1,244
  5. 2025OpenAI GPT-5 System Cardcited by 989
  6. 2023DSPy: Compiling Declarative Language Model Calls into Self-Improving Pipelinescited by 832
  7. 2022Selective Annotation Makes Language Models Better Few-Shot Learnerscited by 243
  8. 2024Large Language Models are Inconsistent and Biased Evaluatorscited by 153
  9. 2024Aligning with Human Judgement: The Role of Pairwise Preference in Large Language Model Evaluatorscited by 149

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2023Large Language Models are biased to overestimate profoundness
  2. 2025Evidence of conceptual mastery in the application of rules by Large Language Models
  3. 2025Language Models Might Not Understand You: Evaluating Theory of Mind via Story Prompting
  4. 2025AIPsychoBench: Understanding the Psychometric Differences between LLMs and Humans
  5. 2026The Grounding Gap: How LLMs Anchor the Meaning of Abstract Concepts Differently from Humans
  6. 2025The Alternative Annotator Test for LLM-as-a-Judge: How to Statistically Justify Replacing Human Annotators with LLMs
  7. 2025DeepQuestion: Systematic Generation of Real-World Challenges for Evaluating LLMs Performance
  8. 2023Psychometric Predictive Power of Large Language Models