Oxford study finds frontier LLM agents fail to coordinate with unfamiliar partners
rohanpaul_ai · x · 2026-11-08
A new Oxford paper finds frontier LLM agents fail to coordinate with unfamiliar partners because each treats its own made-up conventions as the obvious choice — a growing problem as agents from different companies must increasingly work together with no agreed rules.
Testing text and code agents built on GPT-5.2 and Gemini 3 Pro across 9 small cooperative games (e.g., Cat & Dog: safe move guarantees 5 points, signaling convention pays 10 only if matched), paired text agents scored below the safe 5. A short shared "Coordination-friendly definition" led agents to reliably take the safe move.
Practical tip: for agents working with agents you didn't build, instruct them to avoid unstated conventions and prefer actions any partner reads correctly.
More from coding & agent
- Dev benchmarks tiny LLMs on agent tasks: 9B passes 2/5, stepping up to 12B — draginol · 2026-11-08
- Engineer proposes CI/CD could become a handoff to a deployment agent — andreisavu · 2026-11-08
- Dev launches Local AI benchmark scoring models by hardware on agent tasks and speed — draginol · 2026-11-08
- Personal AI dot can now read inbox sign-in codes to log in on your behalf — Dimillian · 2026-11-08
- Dev plans eye-tracker to automatically feed agents visual context — johnlindquist · 2026-11-08
- Lattice: open-source 'Cursor for research papers' bundles Claude Code, Overleaf and citations in one app — minchoi · 2026-11-08