About

Swarat Chaudhuri is a professor specializing in Artificial Intelligence, Formal Methods, Programming Languages, and Compilers. His research focuses on advancing the theoretical foundations and practical applications of these areas, contributing to the development of more reliable, efficient, and secure computing systems. As a faculty member at UT Austin, he is involved in teaching and mentoring students, as well as leading research initiatives that push the frontiers of computer science in these domains.

Research topics

• Computer Science
• Programming language
• Artificial Intelligence
• Machine Learning
• Embedded system
• Theoretical computer science
• Algorithm
• Engineering

Selected publications

• Formal Reasoning Meets LLMs: Toward AI for Mathematics and Verification

  Communications of the ACM · 2026-02-10

  articleOpen access

  Formal reasoning has long been fundamental to various fields of computer science, including AI in its early days. In contrast, large language models (LLMs)—the cornerstone of modern AI—perform reasoning through autoregressive next-word prediction, without grounding their outputs in formal systems. This “informal” approach can learn world knowledge and reasoning patterns from large-scale data without rigid formalization. It offers significantly greater flexibility than traditional formal reasoning and has achieved promising results on many benchmarks. However, LLMs heavily rely on data and do not guarantee the soundness of reasoning. In this article, we highlight recent efforts to integrate modern LLMs with formal methods, an approach that seeks to harness the strengths of both paradigms. Such integration has the potential to lead to major advancements in AI-driven mathematics, formal verification, and the verifiable generation of computer systems.

  PublisherDOI

• Evaluating Agentic Optimization on Large Codebases

  ArXiv.org · 2026-03-16

  articleOpen access

  Large language model (LLM) coding agents increasingly operate at the repository level, motivating benchmarks that evaluate their ability to optimize entire codebases under realistic constraints. Existing code benchmarks largely rely on synthetic tasks, binary correctness signals, or single-objective evaluation, limiting their ability to assess holistic optimization behavior. We introduce FormulaCode, a benchmark for evaluating agentic optimization on large, real-world codebases with fine-grained, multi-objective performance metrics. FormulaCode comprises 957 performance bottlenecks mined from scientific Python repositories on GitHub, each paired with expert-authored patches and, on average, 264.6 community-maintained performance workloads per task, enabling the holistic ability of LLM agents to optimize codebases under realistic correctness and performance constraints. Our evaluations reveal that repository-scale, multi-objective optimization remains a major challenge for frontier LLM agents. Project website at: https://formula-code.github.io

  PublisherOA PDF

• An Improved Last-Iterate Convergence Rate for Anchored Gradient Descent Ascent

  arXiv (Cornell University) · 2026-04-04

  preprintOpen accessSenior author

  We analyze the last-iterate convergence of the Anchored Gradient Descent Ascent algorithm for smooth convex-concave min-max problems. While previous work established a last-iterate rate of $\mathcal{O}(1/t^{2-2p})$ for the squared gradient norm, where $p \in (1/2, 1)$, it remained an open problem whether the improved exact $\mathcal{O}(1/t)$ rate is achievable. In this work, we resolve this question in the affirmative. This result was discovered autonomously by an AI system capable of writing formal proofs in Lean. The Lean proof can be accessed at https://github.com/google-deepmind/formal-conjectures/pull/3675/commits/a13226b49fd3b897f4c409194f3bcbeb96a08515

  PublisherDOI

• Coadaptive Value Alignment

  2026-05-24

  article

  The integration of autonomous agents into human society is a grand challenge for AI. In order to achieve widespread acceptance, agents must conform to the values of people with whom they interact. Current approaches treat the value alignment problem as a unidirectional interaction where the aim is to imbue an agent's actions with human values. Our Coadaptive Value Alignment paradigm acknowledges that human perceptions, expectations, and values continuously evolve in response to agent actions. We conceptualize human-agent interaction as an adaptive loop where the agent actively models and intentionally influences the human's perception, rather than just acting according to static human values. For instance, unlike a traditional agent that simply maximizes speed, an adaptive agent could detect a drop in user trust and strategically sacrifice short-term efficiency to repair the relationship. This perspective transforms value alignment into a multi-agent challenge where all actors must identify and adhere to a shared, implicit social contract. The opportunity to create a virtuous cycle of self-improvement is accompanied by the risk of negative reinforcement, which could result in undesired behaviors. We outline the core framework components, present a research road map for the MAS community, and propose that this dynamic perspective is critical for creating truly collaborative social partners.

  PublisherDOI

• Canopy: Property-Driven Learning for Congestion Control

  2026-04-24 · 2 citations

  preprintOpen access

  Learning-based congestion controllers offer better adaptability compared to traditional heuristics. However, the unreliability of learning techniques can cause learning-based controllers to behave poorly, creating a need for formal guarantees. While methods for formally verifying learned congestion controllers exist, these methods offer binary feedback that cannot optimize the controller toward better behavior. We improve this state-of-the-art via Canopy, a new property-driven framework that integrates learning with formal reasoning in the learning loop. Canopy uses novel quantitative certification with an abstract interpreter to guide the training process, rewarding models, and evaluating robust and safe model performance on worst-case inputs. Our evaluation demonstrates that unlike state-of-the-art learned controllers, Canopy-trained controllers provide both adaptability and worst-case reliability across a range of network conditions.

  PublisherDOI

• Do CFLOBDDs Actually Make Use of Linear Structure?

  ArXiv.org · 2026-05-15

  articleOpen access

  Binary Decision Diagrams (BDDs) are a widely used data structure for efficient Boolean function representation. Context-Free-Language Ordered Binary Decision Diagrams (CFLOBDDs) are a recently introduced hierarchical data structure that can, in the best case, exhibit exponential compression over BDDs and double-exponential compression over decision trees. Roughly speaking, a CFLOBDD is a finite, acyclic, non-recursive hierarchical finite-state machine (HFSM) (with some additional restrictions). In this paper, we investigate the role of \emph{linear structure} in CFLOBDDs -- a property that connects them to Nested-Word Automata (NWAs) and Visibly Pushdown Automata (VPAs) -- and examine whether CFLOBDDs actively exploit this structure beyond their well-studied hierarchical properties. We demonstrate that linear structure, in conjunction with hierarchical structure, plays a crucial role in enabling CFLOBDDs to achieve efficient function compression. Furthermore, we show that removing linearity from CFLOBDDs leads to a significant blowup in representation size, resulting in degraded performance in the domain of quantum-circuit simulation.

  PublisherOA PDF

• An Improved Last-Iterate Convergence Rate for Anchored Gradient Descent Ascent

  arXiv (Cornell University) · 2026-04-04

  articleOpen accessSenior author

  We analyze the last-iterate convergence of the Anchored Gradient Descent Ascent algorithm for smooth convex-concave min-max problems. While previous work established a last-iterate rate of $\mathcal{O}(1/t^{2-2p})$ for the squared gradient norm, where $p \in (1/2, 1)$, it remained an open problem whether the improved exact $\mathcal{O}(1/t)$ rate is achievable. In this work, we resolve this question in the affirmative. This result was discovered autonomously by an AI system capable of writing formal proofs in Lean. The Lean proof can be accessed at https://github.com/google-deepmind/formal-conjectures/pull/3675/commits/a13226b49fd3b897f4c409194f3bcbeb96a08515

  PublisherOA PDF

• CSLib: The Lean Computer Science Library

  ArXiv.org · 2026-02-04

  articleOpen access

  We introduce CSLib, an open-source framework for proving computer-science-related theorems and writing formally verified code in the Lean proof assistant. CSLib aims to be for computer science what Lean's Mathlib is for mathematics. Mathlib has been tremendously impactful: it is a key reason for Lean's popularity within the mathematics research community, and it has also played a critical role in the training of AI systems for mathematical reasoning. However, the base of computer science knowledge in Lean is currently quite limited. CSLib will vastly enhance this knowledge base and provide infrastructure for using this knowledge in real-world verification projects. By doing so, CSLib will (1) enable the broad use of Lean in computer science education and research, and (2) facilitate the manual and AI-aided engineering of large-scale formally verified systems.

  PublisherOA PDF

• Advancing Mathematics Research with AI-Driven Formal Proof Search

  ArXiv.org · 2026-05-21

  articleOpen accessSenior author

  Large language models (LLMs) increasingly excel at mathematical reasoning, but their unreliability limits their utility in mathematics research. A mitigation is using LLMs to generate formal proofs in languages like Lean. We perform the first large-scale evaluation of this method's ability to solve open problems. Our most capable agent autonomously resolved 9 of 353 open Erdős problems at the per-problem cost of a few hundred dollars, proved 44/492 OEIS conjectures, and is being deployed in combinatorics, optimization, graph theory, algebraic geometry, and quantum optics research. A basic agent alternating LLM-based generation with Lean-based verification replicated the Erdős successes but proved costlier on the hardest problems. These findings demonstrate the power of AI-aided formal proof search and shed light on the agent designs that enable it.

  PublisherOA PDF

• FormulaCode: Evaluating Agentic Optimization on Large Codebases

  arXiv (Cornell University) · 2026-03-16

  preprintOpen access

  Large language model (LLM) coding agents increasingly operate at the repository level, motivating benchmarks that evaluate their ability to optimize entire codebases under realistic constraints. Existing code benchmarks largely rely on synthetic tasks, binary correctness signals, or single-objective evaluation, limiting their ability to assess holistic optimization behavior. We introduce FormulaCode, a benchmark for evaluating agentic optimization on large, real-world codebases with fine-grained, multi-objective performance metrics. FormulaCode comprises 957 performance bottlenecks mined from scientific Python repositories on GitHub, each paired with expert-authored patches and, on average, 264.6 community-maintained performance workloads per task, enabling the holistic ability of LLM agents to optimize codebases under realistic correctness and performance constraints. Our evaluations reveal that repository-scale, multi-objective optimization remains a major challenge for frontier LLM agents. Project website at: https://formula-code.github.io

  PublisherDOI

Recent grants

• SHF: Medium: Collaborative Research: Chorus: Dynamic Isolation in Shared-Memory Parallelism

  NSF · $510k · 2011–2015

• SHF: Medium: Collaborative Research: Bridging Automated Formal Reasoning and Continuous Optimization for Provably Safe Deep Learning

  NSF · $500k · 2020–2024

• CAREER: Robustness Analysis of Uncertain Programs: Theory, Algorithms, and Tools

  NSF · $345k · 2011–2017

• SHF: Medium: Collaborative Research: Marrying Program Analysis and Numerical Search

  NSF · $600k · 2012–2018

• CAREER: Robustness Analysis of Uncertain Programs: Theory, Algorithms, and Tools

  NSF · $160k · 2010–2012

Frequent coauthors

• Lydia E. Kavraki

  23 shared

• Işıl Dillig

  The University of Texas at Austin

  20 shared

• Chris Jermaine

  19 shared

• Abhinav Verma

  18 shared

• Neil T. Dantam

  Colorado School of Mines

  15 shared

• Thomas Reps

  14 shared

• Yisong Yue

  California Institute of Technology

  13 shared

• Greg Durrett

  12 shared

Education

• Ph.D., Computer Science

  University of California, Berkeley

  2002

• M.S., Computer Science

  University of California, Berkeley

  1998

• B.S., Electrical Engineering

  University of Calcutta

  1996

Awards & honors

• NSF CAREER award
• Google Research Award
• ACM SIGPLAN John Reynolds Doctorate Dissertation Award
• multiple distinguished paper awards