CS 839 · Fall 2026

AI Agents for Science

Preliminary schedule
Meeting time. Tuesdays and Thursdays, 9:30–10:45 AM, in Mechanical Engineering 1164. Readings, presentation assignments, and project deadlines will be added before the semester.
DateSessionFocus
Part I. Foundations and capabilities of LLM Agents
Thu, Sept. 3Lecture 1: Introduction: AI Agents for ScienceOverview. Agents, LLMs, and scientific discovery
Tue, Sept. 8Lecture 2: From Language Models to Agents Instruction following, reasoning models, and agent loops
Thu, Sept. 10Lecture 3: Agent Architectures and Frameworks Orchestration, state, control flow, debugging
Tue, Sept. 15Lecture 4: Tools, Environments, and Agent–Computer InterfacesFunction calling, code execution, scientific software, and error recovery
Thu, Sept. 17Lecture 4 (continued): Tools, Environments, and Agent–Computer InterfacesContinued from Sept. 15
Tue, Sept. 22Lecture 5: Planning and Task DecompositionReAct, search, reflection, verification, and long-horizon tasks
Thu, Sept. 24Lecture 6: Memory, Context, and RetrievalMemory systems, retrieval, updates, forgetting, and evaluation
Part II. Scientific applications
Tue, Sept. 29Lecture 7: Agents for Scientific Literature and Evidence Synthesis Search, citation verification, systematic reviews, and research gaps
Thu, Oct. 1Lecture 8: Coding Agents for Science and BeyondCode, repos, tools, environments, tests, verification
Tue, Oct. 6Lecture 9: Agents for Hypothesis Generation and Experimental DesignHypotheses, experiment selection, active learning + Presentation information
Thu, Oct. 8Lecture 10: Auto-Laboratories and Autonomous ExperimentationRobotics, closed-loop experimentation, costly actions, and approval
Part III. Evaluation, learning, and open frontiers
Tue, Oct. 13Lecture 11: Evaluation and Benchmarks for AgentsValidity, baselines, reproducibility, cost, and discovery
Thu, Oct. 15Lecture 12: Training and Improving AgentsTrajectory data, process supervision, reinforcement learning, and credit assignment
Tue, Oct. 20Lecture 13: Multi-Agent System & Scientific DiscoverySpecialized roles, debate, coordination, and failure
Thu, Oct. 22Lecture 14: Human–Agent Interaction, Safety, and Scientific IntegrityOversight, sandboxing, credentials, evidence, attribution, and novelty
Tue, Oct. 27Lecture 15: Scientific Case Studies and Open FrontiersBiology, chemistry, materials, mathematics, climate, and new applications
Part IV. Student presentations
Thu, Oct. 29Student Presentation 1TBA
Tue, Nov. 3Student Presentation 2TBA
Thu, Nov. 5Student Presentation 3TBA
Tue, Nov. 10Student Presentation 4TBA
Thu, Nov. 12Student Presentation 5TBA
Tue, Nov. 17Student Presentation 6TBA
Thu, Nov. 19Student Presentation 7TBA
Tue, Nov. 24Student Presentation 8TBA
Thu, Nov. 26No classThanksgiving recess
Tue, Dec. 1Student Presentation 9TBA
Thu, Dec. 3Student Presentation 10TBA
Tue, Dec. 8Course Synthesis and Project DiscussionWhat have we learned, and what is next?