Google's ScientistTwo uses autonomous multi-agent AI to produce publishable, verified research papers
burny_tech · x · 2026-10-07
Google Research demoed ScientistTwo at COLM, a fully autonomous multi-agent framework for scientific discovery, with the paper now on arXiv.
- Given an initial problem, it establishes SOTA baselines, formulates novel hypotheses, and coordinates specialized agents through an end-to-end discovery cycle with no human intervention
- It runs experiments across diverse datasets and metrics, iterates via automated ablation studies, and validates findings with a closed-loop simulated peer-review rebuttal engine
- Benchmarked against papers accepted at ICLR, ICML, and NeurIPS, the authors claim it autonomously generates expert-level, publishable papers plus fully verified executable codebases, outperforming human state of the art
Authors include Jaehyun Nam, Jinsung Yoon, and Rui Meng; a live demo was held at Google's COLM booth.
More from Research
- Looped transformers are trending again, tracing roots back to a 2016 paper — natanielruizg · 2026-10-07
- Fixed token codes suffice: 1.7B LM trains without a trainable input embedding table — A. Bochkov · 2026-10-07
- EmbeddingGemma 2 hands-on: 740M multimodal embeddings for search and RAG, runnable on a free T4 — Prompt Engineering · 2026-10-07
- Isomorphic, DeepMind and Meta join DOE-NIH partnership to build an AI model of the cell — snikolov · 2026-10-07
- Bi-manual mobile UMI demo unlocked for robot manipulation data collection — neurosp1ke · 2026-10-07
- Researcher presents Meta-Harness and Combee at COLM 2026, seeks industry roles — Kangwook_Lee · 2026-10-07