An Oxford paper finds that cutting-edge LLM agents struggle to collaborate with unfamiliar partners via zero-shot learning
A research paper from Oxford's BOLD team tested zero-shot collaboration (ZSC) of LLM agents in nine small cooperative games using text and code agents based on models such as GPT-5.2 and Gemini 3 Pro.
Continue looking at this topic
OpportunitiesAt over 60, he built a website inspector using AI and created a set of universal templates ↗Related contentWhat is Google Antigravity? Distinguish between the development platform and the Gemini model ↗News and AnalysisCan OpenCode Browse enable agents to control real browsers? ↗Open-source projectHow to choose the right Claude Code terminal agent? ↗Return to original text
A short introduction is retained here; the full text from the source website is not reposted. For specific wording and subsequent updates, please refer to the original article.
Open AIHOT original