Mechanist is an agentic system that uses AI as a scientific instrument for the autonomous discovery of mechanisms underlying AI intelligence, and develops a mechanism theory of belief, revealing how models represent world knowledge, form beliefs, infer the beliefs of others, and how these mechanisms emerge during pretraining.
Abstract
AI models have achieved remarkable success across diverse domains, yet the mechanisms underlying their capabilities and the risks they may pose remain poorly understood. As AI development becomes faster and increasingly automated, mechanistic exploration remains largely manual, widening the gap between what models can do and our ability to understand and control them. To bridge this gap, we introduce Mechanist, an agentic system that uses AI as a scientific instrument for the autonomous discovery of mechanisms underlying AI intelligence. To support autonomous mechanistic discovery, we construct an interpretability-focused knowledge graph of approximately 13,000 papers and integrate it with a multidisciplinary database of 43 million papers spanning 26 fields. We further curate a library of 32 foundational methods for mechanism analysis, causal intervention, and validation. Compared with Claude Code and existing AI-scientist systems, Mechanist generates more valuable mechanism hypotheses and executes experiments more reliably. Mechanist also demonstrates a progression from discovering model behaviors to explaining and controlling AI models. Specifically, Mechanist first uncovers a counterintuitive safety risk in scientific laboratories, showing that unsafe traits can transfer across modalities through apparently safe training data. Mechanist then develops a mechanism theory of belief, revealing how models represent world knowledge, form beliefs, infer the beliefs of others, and how these mechanisms emerge during pretraining. Finally, Mechanist translates these mechanistic insights into practical interventions that improve model performance across diverse scenarios and steer scientific foundation models toward generating DNA sequences with specified properties.
Artificial intelligence in science is undergoing a foundational change. Rather than serving as a passive analytical instrument — classifying images, predicting structures, spotting patterns — AI systems are beginning to act as autonomous research collaborators. These systems, built on large language models and tool-integrated architectures, can reason about experimental design, formulate strategies, execute multi-step workflows, and refine their approaches from empirical feedback. Often called “AI Scientists,” they participate across the full research lifecycle, from the seed of a hypothesis through to a draft manuscript. This article examines the emerging paradigm of agentic AI for scientific discovery. It traces the conceptual shift from tools to agents, lays out a six-stage workflow spanning literature synthesis to manuscript generation, and reviews practical systems in chemistry, equation discovery, materials science, and general machine learning research. A central concern of the analysis is the verification crisis — the growing gap between what these systems can produce and what they can prove. We compile quantitative evidence on failure rates, analyse competing frameworks for trustworthy agentic science (Chain-of-Evidence, Audit-Closed protocols, FEV, and structural FDR enforcement), and propose actionable standards for rigorous validation. The article closes with an assessment of the field’s limitations and a set of priority directions for making agentic science trustworthy at scale.
Alexander Taktakidze· Longevity Horizon· 0 citations
Recent advances in foundation models have transformed AI for Science, enabling remarkably accurate predictive performance across domains ranging from protein folding to weather forecasting. Yet prediction alone does not constitute scientific discovery. Scientific understanding depends on uncovering the reusable explanatory mechanisms that generate observations, whereas contemporary machine learning remains fundamentally organised around predictive mappings rather than explanatory structure. In this paper, we argue that scientific discovery is fundamentally a problem of knowledge organisation. To this end, we introduce Mechanistic World Models, a new design paradigm that places reusable mechanisms at the centre of representation, computation and learning. Drawing on insights from the philosophy of science, we derive the computational capabilities required for discovery, identify the design principles and inductive pressures that encourage explanatory knowledge to emerge, and formalise the anatomy of a mechanism-centric world model. Finally, we show how diverse research directions including mechanistic interpretability, causal representation learning, equation discovery and modular architectures capture complementary ingredients of this paradigm while lacking a unified framework. We propose Mechanistic World Models as a conceptual foundation and computational blueprint for moving AI beyond predictive forecasting towards autonomous scientific discovery.
Automation is transforming scientific discovery by enabling systematic exploration of complex hypotheses. Large language models (LLMs) perform well across diverse tasks and promise to accelerate research, but often struggle with logical structures. Here, we present a framework for biological discovery integrating LLM-based agents with laboratory automation, guided by logical scaffolds incorporating symbolic relational learning, structured vocabularies and experimental constraints. This integration improves coherence and reliability in automated workflows. We couple this AI-driven approach to automated cell-culture and metabolomics platforms, enabling integrated hypothesis validation and refinement, yielding a flexible discovery system. The system identified novel interactions in Saccharomyces cerevisiae, including glutamate-induced growth inhibition in spermine-treated cells and aminoadipate's partial rescue of formic-acid stress. All hypotheses, experiments and data are captured in a graph database employing controlled vocabularies. Existing ontologies are extended, and a novel representation of scientific hypotheses is presented using description logics. This work demonstrates the potential for a reliable machine-driven discovery process in systems biology.
Daniel Brunnsåker, Alexander H. Gower, Prajakta Naval et al.· Journal of the Royal Society...· 2 citations
We present a survey of the past and future of AI Scientists: machines capable of automating science. AI Scientists can originate hypotheses, deduce their consequences, design and execute experiments, interpret their results, and revise their beliefs. Such systems are integrated scientific agents, connected to the literature, formal knowledge, mathematical models, simulations, data-analysis systems and physical laboratories. Adam was the first machine to make novel scientific discoveries through cycles of hypothesis formation and physical experimentation. Eve established the architecture of the modern self-driving laboratory. Foundation models, autonomous agents and laboratory robotics now make it possible to build systems far more general than either Adam or Eve. The central problem is no longer whether individual components of science can be automated. They can. The problem is integration. AI Scientists must combine neural learning with logic, probability, mathematics, causal reasoning, simulation, experimental design, robotics and formal scientific records. AI Scientists have the potential to transform science: to make science faster, cheaper, more systematic and more reproducible. AI Scientists could investigate systems too complicated for unaided human science, and enable thousands of AI scientists to work together on single problems. The Nobel Turing Challenge sets the goal of developing by 2050 AI systems capable of automating Nobel-quality discoveries. Progress is ahead of schedule. When we succeed it will create a new form of science and transform the world.
Large language models and foundation models are increasingly embedded in reasoning systems that plan, invoke tools, use memory, gather evidence, and iteratively refine their outputs. The second KDD Day on AI Reasoning brings together researchers and practitioners from academia and industry to examine how these systems can be made more capable, reliable, interpretable, and efficient. The program spans scientific discovery, human-centered interaction, software engineering, time-series analysis, deep research, computer use, and inference infrastructure. Across these domains, the day highlights shared challenges: grounding decisions in evidence, designing effective feedback and verification mechanisms, evaluating open-ended behavior, managing test-time computation, and preserving meaningful human control. Through keynote and invited presentations, the event provides a forum for connecting advances in models, agents, data, systems, and applications, and for identifying research directions toward trustworthy next-generation reasoning systems.
Jun Huan, James Caverlee, Lei Li et al.· Proceedings of the 32nd ACM...· 0 citations
Artificial intelligence is driving a revolution in scientific discovery, accelerating everything from hypothesis generation to mathematical theorem proving. However, this rapid acceleration is creating a systemic challenge: traditional human peer review cannot scale to match the influx of AI-assisted science. Ultimately, to resolve this tension, we must also deploy AI to accelerate the verification and review process itself. To frame the discussion around this transition, we propose a taxonomy consisting of four progressive levels of AI-human collaboration in scientific evaluation, and discuss various trade-offs involved with each. As a step toward this future, we introduce the Paper Assistant Tool (PAT), an agentic AI framework built for deep scientific review and verification. PAT ingests full scientific manuscripts and produces a comprehensive evaluation, checking theoretical results, validating experiments, suggesting improvements, and identifying potential flaws. By utilizing inference scaling techniques, PAT is able to identify deeper issues than a single model call alone, achieving a 34% improvement over zero-shot recall on mathematical errors in the SPOT benchmark. Pilot deployments of PAT as a pre-submission tool for authors at two major Computer Science conferences -- STOC and ICML -- demonstrate its ability to identify critical errors and suggest substantive improvements to research papers. By catching errors early, PAT eases the cognitive burden placed on referees, while preserving their control over the outcomes of the review process.
Rajesh Jayaram, Drew Tyler, David P. Woodruff et al.· arXiv.org· 0 citations
Known for his clear and elegant writing style, Bertsekas shaped fields from control and optimization to large-scale computation and artificial intelligence.
