Reproduce “A Framework for Context Enrichment and Selective Robustness in LLM Social Agents” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Mattia Greiche, Reihaneh Rabbany, Maximilian Puelma Touzel
The authors test whether language-model agents respond to social-world changes while remaining stable across equivalent prompt formats in a hospital supply-transfer simulation. They compare five context conditions across two worlds and four formats. For GPT-5.6 Sol, descriptive decision grounding was associated with greater world responsiveness and lower format sensitivity than generic social background. Terra remained format-sensitive in some grounded conditions; this pilot covers one setting and two models.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.