ai-scientist

The paper proposes AgenticTuner, an agentic pipeline using an LLM as the optimizer to calibrate cosmological simulations by leveraging human-readable priors like codebases and documentation, achieving competitive calibration performance at a fraction of the CPU cost of Bayesian optimization.

Google Cloud AI Research and Peking University jointly released PaperBanana (later renamed PaperVizAgent) — an open-source framework that uses multi-agent collaboration to automatically generate publication-quality academic illustrations (method flowcharts, architecture diagrams, statistical plots). Its pipeline of "retrieve reference images → plan content → adjust style → generate → self-critique and iterate" lowers the barrier to creating paper figures, accompanied by the evaluation benchmark PaperBananaBench with 292 test cases from NeurIPS 2025, outperforming existing baselines in faithfulness, conciseness, readability, and aesthetics.

The paper introduces EvoSim, a self-evolving multi-agent AI scientist that autonomously constructs and refines physics-based models using experimental discrepancies, evaluated on lithium-metal battery modeling where it surpasses human-expert accuracy.

A roundup of major AI developments: Trump announced the federal Super Intelligence Force with an AI Czar, ex-Anthropic researcher Jacob Coxon warned humanity is likely to lose control of AI, and research labs reported surging AI-driven scientific output. The author argues SI is now federal policy even as labs admit they cannot control these systems.

Researchers at Chalmers University of Technology developed a closed-loop AI scientist that generates biological hypotheses, designs and runs experiments via lab robots, and learns from the results, testing it on yeast by producing 1,933 candidate hypotheses.

This paper presents AIMC, a visual analytics framework for human oversight of autonomous scientific discovery, enabling monitoring and understanding of AI-generated research artifacts.

This podcast episode explores the concept of AI scientists, which combine large language models with automated laboratories to design and conduct experiments. Ant Rowstron from ARIA discusses the emerging architecture and ongoing research in the field.

Jeff Dean, former Google Chief Scientist, tearfully discusses the decision to leave Google after 27 years and start an independent startup, expressing deep feelings for his time at Google.

This article recommends paying attention to AI Scientist and discusses how research agents can learn from failures by analogizing to fuzz testing, thereby mapping the unknown and guiding subsequent experiments.

OmniScientist is an end-to-end omni-modal AI scientist that performs multidisciplinary research directly from heterogeneous raw evidence using autonomous agents and lifecycle-wide perception. Evaluated on 36 real-data cases, it improves evidence-grounded discovery across diverse scientific modalities.

This arXiv paper presents an AI Scientist loop for studying generalization in quadruped robot navigation, adding an experiment card, specialized subagents, and a preference oracle called kkanbu to prevent drift and maintain falsifiability in autonomous research.

This paper proposes a benchmarking protocol using automated multi-model LLM review to evaluate AI Scientist systems, comparing frameworks like Sakana AI, CycleResearcher, and Data-to-Paper, and finds that FARS benchmark papers significantly outperform other systems.

Sakana AI has released AI Scientist-v2, an end-to-end automated research system that can autonomously go from generating research hypotheses to writing papers, and has been accepted by the ICLR2025 Workshop after peer review.

This paper introduces Xcientist, a research harness that externalizes AI-driven scientific research synthesis and validation into inspectable, contract-governed processes to ensure accountability and traceability.

Researchers at MIT present a paper on self-evolving AI scientists that can discover and adapt their own scientific vocabulary, using a categorical framework to mathematically quantify genuine novelty and separate discovery from mere search or retrieval.

Harvard University's AutoScientists proposes a decentralized multi-agent team approach, allowing multiple agents to share experimental status, automatically form teams, and review research plans, significantly outperforming existing methods on multiple benchmarks.

A new paper from Meta, Stanford, and Google introduces AutoResearchClaw, which improves automated research by integrating failure recovery, debate, and selective human input. It outperforms AI Scientist v2 by 54.7% on ARC-Bench and reveals that autonomy is enhanced when constrained by process rather than given unlimited freedom.

A study evaluates frontier models' ability to forecast scientific progress across 4,760 events, finding they can identify plausible directions but cannot reliably predict outcomes or timelines, with systematic overconfidence.

A comprehensive open-source collection of 138 scientific agent skills that transform AI coding assistants like Claude Code and Codex into AI scientists, covering biology, chemistry, medicine, and more, with integration of over 100 scientific databases and specialized Python packages.

This paper presents AI CFD Scientist, an open-source AI agent for computational fluid dynamics that autonomously discovers physics corrections using vision-language verification and code modification, outperforming general AI scientists on CFD tasks.