ROTATE Ad Hoc Teamwork Training Paper Wins NeurIPS 2026 Spotlight
cuijiaxun · x · 2026-09-25
- The ROTATE paper (Regret-driven Open-ended Training for Ad Hoc Teamwork) by Caroline Wang, Jiaxun Cui, Peter Stone et al. is accepted as a NeurIPS 2026 Spotlight.
- It tackles Ad Hoc Teamwork (AHT): collaborating with unseen partners. Prior methods use a fixed teammate population in a two-stage pipeline, limiting behavior coverage.
- ROTATE reformulates AHT as open-ended learning between the AHT agent and an adversarial teammate generator, using regret to alternate between improving the agent and generating teammates that probe its weaknesses.
- On Overcooked and Level-Based Foraging, ROTATE substantially outperforms baselines on unseen teammate sets; paper and code are open.
More from Research
- Cameron Wolfe publishes complete guide tracing RL for LLMs from VPG to GRPO variants — cwolferesearch · 2026-09-25
- Question's Gambit tops BrowseComp-Plus recall with 96.6% using only BM25 — CShorten30 · 2026-09-25
- ICLR page-limit tip: use \textbf instead of \paragraph to save space — jindong_wang92 · 2026-09-25
- AutoScientists NeurIPS paper: self-organizing AI research teams hit 74.4 percentile on BioML-Bench — marinkazitnik · 2026-09-25
- Does the curse of multilinguality have to exist in theory? Embedding-space study — mdredze · 2026-09-25
- From VPG to GRPO: the clean evolution of RL algorithms behind LLM training — cwolferesearch · 2026-09-25