Reproduce “Bad Apples in the Kitchen: How LLM Teams Detect, Correct, and Calibrate Against Their Misaligned Members” 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
Dongseok Heo, Junghwan Kim, Bongwon Suh
The authors study recovery in real-time Overcooked teams that act while communicating. They compare weak monitoring, which prompts bottleneck queries, with strong monitoring, which also requires attribution, diagnosis, and a corrective directive. Across 720 episodes, attribution to Player 2 occurred in 88–97%, but combined attribution and diagnosis in 31–61%. Strong monitoring improved Disruptor performance but reduced troll-free performance in this custom environment.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.