Reproduce “Read the Room Before Your Model Generates” 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
Muyao Kong, Zhihao Lin
Scenario-Actor selects a prompt-level temperature from an inferred communicative scenario. Scenarize represents prompts; Actout combines estimates across five axes. On AgentSense, its score of 0.8015 versus 0.7801 for fixed temperature was not statistically significant. It significantly exceeded class-conditioned lookup, scoring 0.8015 versus 0.7536. Tested on one model family, it scored 65.28% on GSM8K, below greedy decoding at 77.18%.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.