Research map / Large language models / Agents, code and math
Mind the Gap: The Divergence Between Human and LLM-Generated Tasks
Yi-Long Lu, Jiajun Song, Chunhui Zhang, Wei Wang
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 changed | Where | Measured | Effect |
|---|---|---|---|
| condition GPT-4o on matched human psychological profile (values, cognitive style) (ablation) | GPT-4o task generation · Task theme distribution | share of Relaxation & Household tasks: 3% (raw) vs 5% (matched) vs 32% (humans) | No clear effect |
| condition GPT-4o on matched human profile (ablation) | GPT-4o task generation · Multiplayer tasks (another person present) | share of multiplayer tasks: 29% -> 24% (humans 58%) | No clear effect |
| higher Openness to Change value in human participants | human task generation · Human task ratings | task novelty rating: b = 0.152, p = 0.017 | Improved |
| higher Openness to Change value in human participants | human task generation · Human task ratings | task fun rating: b = 0.146, p = 0.022 | Improved |
| self-interest value dimension as predictor | human task generation · Human task ratings | novelty and fun | No clear effect |
| more systematic thinking style (TWS x environmental complexity interaction) | human task generation · Human task ratings | task fun: interaction b = -0.140, p = 0.009 | Mixed |
| more systematic thinking style (TWS x environmental complexity interaction) | human task generation · Human task diversity | task diversity (cosine distance): interaction b = -0.215, p = 0.006 | Mixed |
| use GPT-4o (matched) instead of humans | task generation · Rated mental demand | mental demand rating: z = -11.64, pFDR < 0.001 | Hurt |
| use GPT-4o (matched) instead of humans | task generation · Rated physical demand | physical demand rating: z = -8.40, pFDR < 0.001 | Hurt |
| use GPT-4o (matched) instead of humans | task generation · Rated novelty | novelty rating: z = -5.63, pFDR < 0.001 | Improved |
| use GPT-4o (matched) instead of humans | task generation · Rated fun | fun rating: z = -2.73, pFDR = 0.010 | Improved |
| use GPT-4o (matched) instead of humans | task generation · Body part engagement | arms and lower body engagement: arms z = -5.43; lower body z = -7.93 | Hurt |
Closest papers by meaning
- Language Model Goal Selection Differs from Humans' in a Self-Directed Learning Task (NeurIPS 2026 · similarity 0.91)
- Open Models, Closed Minds? On Agents Capabilities in Mimicking Human Personalities through Open Large Language Models (AAAI 2025 · similarity 0.90)
- LLM Agents Can Be Choice-Supportive Biased Evaluators: An Empirical Study (AAAI 2025 · similarity 0.88)
- Embodied Agent Interface: Benchmarking LLMs for Embodied Decision Making (NeurIPS 2024 · similarity 0.88)
- Simulating Society Requires Simulating Thought (NeurIPS 2025 · similarity 0.88)
- MindForge: Empowering Embodied Agents with Theory of Mind for Lifelong Cultural Learning (NeurIPS 2025 · similarity 0.87)
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.
