Introduces HumanStudy-Bench to evaluate LLM agents against 12 replicated human behavioral studies, finding agent design affects alignment more than model scale with polarized outcomes.
Given the heterogeneity of the studies—which span diverse cognitive and social domains—we treat each study as an independent unit of capability
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CY 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Validated Hypotheses as a Lens for Human-Likeness Evaluation in AI Agents
Introduces HumanStudy-Bench to evaluate LLM agents against 12 replicated human behavioral studies, finding agent design affects alignment more than model scale with polarized outcomes.