Publications
Our teams aspire to make discoveries that impact everyone, and core to our approach is sharing our research and tools to fuel progress in the field.
Our teams aspire to make discoveries that impact everyone, and core to our approach is sharing our research and tools to fuel progress in the field.
Sort By
1 - 15 of 11379 publications
From Correctness to Collaboration: A Human-Centered Taxonomy of AI Agent Behavior in Software Engineering
Extended Abstracts of the 2026 CHI Conference on Human Factors in Computing Systems (CHI EA ’26), ACM, New York, NY, USA (2026)
Preview abstract
The ongoing transition of Large Language Models in software engineering from code generators into autonomous agents requires a shift in how we define and measure success. While models are becoming more capable, the industry lacks a clear understanding of the behavioral norms that make an agent effective in collaborative software development in the enterprise. This work addresses this gap by presenting a taxonomy of desirable agent behaviors, synthesized from 91 sets of user-defined rules for coding agents. We identify four core expectations: Adhere to Standards and Processes, Ensure Code Quality and Reliability, Solve Problems Effectively, and Collaborate with the User. These findings offer a concrete vocabulary for agent behavior, enabling researchers to move beyond correctness-only benchmarks and design evaluations that reflect the realities of professional software development in large enterprises.
View details
ToolGrad: Efficient Tool-use Dataset Generation with Textual "Gradients"
Kohei Uehara
Haoyu Zhang
Lin Gu
Zheng Xu
Tatsuya Harada
ACL 2026 (2026)
Preview abstract
Prior work synthesizes tool-use LLM datasets by first generating a user query, followed by complex tool-use annotations like depth-first search (DFS). This leads to inevitable annotation failures and low efficiency in data generation. We introduce ToolGrad, an agentic framework that inverts this paradigm. ToolGrad first constructs valid tool-use chains through an iterative process guided by textual "gradients", and then synthesizes corresponding user queries. This "answer-first" approach led to ToolGrad-500, a dataset generated with more complex tool use, lower cost, and almost 100% pass rate. Experiments show that ToolGrad models outperform those trained on expensive baseline datasets and proprietary LLMs.
View details
Preview abstract
This article delves into how Google Site Reliability Engineers (SREs) leverage Gemini 3 and the Gemini CLI to aggressively reduce Mean Time to Mitigation (MTTM) during real-world outages. By focusing on the SRE motto of "Eliminate Toil," the article walks through a simulated incident, demonstrating how an agentic CLI acts as a human-in-the-loop copilot across the entire incident lifecycle: from initial paging and investigation, through safe, tool-driven mitigation and root cause analysis, to automated postmortem generation and action item filing. This direct integration of Gemini's reasoning capabilities with operational data and internal tools creates a virtuous cycle where past incident learnings continuously inform and improve future solutions.
View details
Towards AI as a Collaborative Partner: A Taxonomy of AI Agent Behavior in Software Engineering
Proceedings of the 3rd ACM International Conference on AI-Powered Software (AIware '26), ACM, Montreal, QC, Canada (2026) (to appear)
Preview abstract
The ongoing transition of Large Language Models (LLMs) in software engineering from one-shot code generators into agentic partners requires a shift in how we define and measure success. While models are becoming more capable, the industry lacks a clear understanding of the behavioral norms that make an interactive software engineering (SWE) agent effective in collaborative software development in the enterprise. This work addresses this gap by presenting a taxonomy of desirable SWE agent behaviors, synthesized from 91 sets of developer-defined rules for SWE agents and validated through interviewing 15 experienced professional developers. In this taxonomy, we identify four core expectations: Adhere to Standards and Processes, Ensure Code Quality and Reliability, Solve Problems Effectively, and Collaborate with the Developer. These findings offer a concrete vocabulary for aligning SWE agent behavior with developer preferences, enabling researchers and practitioners to move beyond correctness-only benchmarks and start designing evaluations that reflect the socio-technical nature of professional software development in enterprises.
View details
XProf: An Open, Scalable and Extensible Profiling System for the Modern ML Stack
Robert Hundt
Naveen Kumar
Jose Baiocchi Paredes
Scott Goodson
Kelvin Le
Yin Zhang
Kan Cai
Sai Ganesh Bandiatmakuri
Yogesh SY
Ani Udipi
Vikas Aggarwal
Ninth Conference on Machine Learning and Systems (2026)
Preview abstract
Optimizing Large Models across thousands of accelerators requires deep system expertise. To address modern machine learning (ML) optimization needs, we present XProf, the ML profiler for the OpenXLA ecosystem. XProf delivers actionable optimization suggestions and in-depth performance analysis, empowering ML researchers and framework users to improve efficiency without specialized systems knowledge. XProf provides a unified, full-stack view of both host (CPU) and device (accelerator - TPUs/GPUs) performance, leveraging tools like the Roofline Model for comprehensive analysis. XProf’s distributed architecture is designed to monitor thousands of chips with minimal workload overhead (<1%). This architecture is made pluggable through the open-source PJRT C API extension, which has facilitated its adoption by third-party accelerator vendors. XProf has been instrumental in achieving significant efficiency gains at Google and winning MLPerf submissions. This paper presents the design and architecture of XProf, showcases its differentiating tools and capabilities, and highlights its impact within Google and across the industry as a state of the art ML profiler. XProf is available as part of the OpenXLA project at https://github.com/openxla/xprof.
View details
Gaze Target Estimation Anywhere with Concepts
Xu Cao
Houze Yang
Vipin Gunda
Inki Kim
Jim Rehg
Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (2026)
Preview abstract
Estimating human gaze targets in-the-wild is a formidable challenge. Existing computer vision algorithms rely on brittle, multi-stage pipelines that require explicit inputs like head bounding boxes and human pose, causing initial detection errors to cascade and lead to system failure. To overcome this, we introduce the \textbf{Promptable Gaze Target Estimation (PGE)} task, a new end-to-end, concept-driven paradigm. PGE conditions gaze prediction on flexible user text or visual prompts (e.g., "the boy in the red shirt" or "person in point [0.52, 0.48]") to identify a specific subject's target, which eliminates the rigid dependency on intermediate localization cues. We develop a scalable data engine to generate \textbf{Gaze-Co}, a dataset and benchmark of 120K high-quality, prompt-annotated image pairs. We also propose \textbf{AnyGaze}, the first model designed for PGE. AnyGaze uses a Transformer-based detector to fuse features from frozen encoders and simultaneously solves subject localization, in/out-of-frame presence, and gaze target heatmap estimation. AnyGaze achieves state-of-the-art performance on standard gaze target estimation benchmarks, setting a strong baseline for this new problem even on a difficult out-of-domain, real-world clinical dataset. We will open-source the AnyGaze model and the Gaze-Co benchmark.
View details
Preview abstract
Voice activity detection (VAD) plays a vital role in enabling applications such as speech recognition. We analyze the impact of window size on the accuracy of three VAD algorithms: Silero, WebRTC, and Root Mean Square (RMS) across a set of diverse real-world digital audio streams. We additionally explore the use of hysteresis on top of each VAD output. Our results offer practical references for optimizing VAD systems. Silero significantly outperforms WebRTC and RMS, and hysteresis provides a benefit for WebRTC.
View details
JAXBench: Benchmarking Autonomous TPU Kernel Optimization
Vijay Janapa Reddi
Charles Hong
Arya Tschand
Julian Walker
Suvinay Subramanian
Shangkun Wang
Sethu Sankaran
Nina Cai
2026
Preview abstract
Evaluation benchmarks have driven progress in automated kernel optimization, yet existing suites target GPUs exclusively. We present JAXBench, a TPU-native benchmark for AI-generated kernel optimization on Google Cloud TPUs. JAXBench comprises 50 JAX workloads, including 17 production LLM operators extracted from architectures in the public MaxText library such as Llama-3.1, DeepSeek-V3, Mixtral, Mamba-2, and AlphaFold2, and 33 fused operator sequences adapted from KernelBench. Eight of the 17 production operators ship with hand-optimized Pallas TPU kernels from the public Tokamax library, whose block sizes we tune via grid search, establishing strong reference baselines. We evaluate one-shot generation, iterative coding agents, the same iterative loop with TPU documentation injected, and a TPU-enabled Autocomp configuration augmented with the same TPU-specific documentation. Across the full 50-benchmark suite with Gemini 3 Flash, best-of-N solves 13/50 benchmarks at 1.01x geomean and iterative refinement reaches 32/50 at 1.18x. Injecting TPU documentation lifts iterative refinement to 48/50 at 1.28x and raises per-sample correctness from 5.8% to 37.3%. Autocomp solves 45/50 but converts those correct kernels into 1.36x geomean with 76% of benchmarks beating XLA. On the 8 hand-tuned references, Autocomp reaches 1.60x against the XLA baseline, showing strong performance against the 2.08x Tokamax geomean but trailing on the specialized paged and ragged attention operators. A Gemini 3.1 Pro ablation lifts Autocomp to 49.1% per-sample correctness and a 3.13x geomean. High-quality TPU kernel generation remains open, and we release the benchmark, profiling harness, and baseline results to support reproducible research.
View details
Preview abstract
Conformal prediction provides rigorous uncertainty guarantees for model outputs but can produce prohibitively large prediction sets in structured domains such as routing, planning, or sequential recommendation. We introduce \emph{graph-based conformal compression}, a framework for constructing compact subgraphs that preserve the statistical validity of conformal prediction while reducing structural complexity. We study a formulation that selects a smallest subgraph capturing a prescribed fraction of conditional probability mass, and reduce to a weighted version of densest $k$-subgraphs in hypergraphs, in the regime where the subgraph has a large fraction of edges. We design efficient approximation algorithms that achieve constant factor coverage and size trade-offs. Our results highlight an algorithmic regime, distinct from classical densest-$k$-subgraph hardness settings, where the problem can be approximated efficiently, bridging conformal prediction with combinatorial graph compression. We finally validate our algorithmic approach on synthetic and real-world instances of trip planning and navigation, showing in each case that our approach handily beats natural baselines.
View details
Who Controls the Curriculum for AI? The Limits of Participatory Design for Educational AI
Learning Under Algorithmic Conditions, University of Minnesota Press (2026)
Preview abstract
Participatory design is a long-standing effort to shift control over technology design from technologists to users and communities impacted by technologies. For educational AI, this means involving students, families, teachers, and other stakeholders in shaping the design of AI systems. While promising, in this article, I situate the recent calls for participatory design of educational AI systems within a different historical tradition—that of contests over local control of educational curricula. I argue that approaches that attempt to steer the design and development of educational AI through participatory methods may inadvertently reproduce the history of political contestation of educational curricula, in ways that may privilege the most powerful communities, rather than those inequitably impacted. What might it look like to treat participatory AI design as a site for political contestation? How might these approaches avoid reproducing the same majoritarian tendencies that led to educational inequities in the first place?
View details
Exponential quantum advantage in processing massive classical data
Haimeng Zhao
Alexander Zlokapa
John Preskill
Hsin-Yuan (Robert) Huang
arXiv:2604.07639 (2026)
Preview abstract
Broadly applicable quantum advantage, particularly in classical data processing and machine learning, has been a fundamental open problem. In this work, we prove that a small quantum computer of polylogarithmic size can perform large-scale classification and dimension reduction on massive classical data by processing samples on the fly, whereas any classical machine achieving the same prediction performance requires exponentially larger size. Furthermore, classical machines that are exponentially larger yet below the required size need superpolynomially more samples and time. We validate these quantum advantages in real-world applications, including single-cell RNA sequencing and movie review sentiment analysis, demonstrating four to six orders of magnitude reduction in size with fewer than 60 logical qubits. These quantum advantages are enabled by quantum oracle sketching, an algorithm for accessing the classical world in quantum superposition using only random classical data samples. Combined with classical shadows, our algorithm circumvents the data loading and readout bottleneck to construct succinct classical models from massive classical data, a task provably impossible for any classical machine that is not exponentially larger than the quantum machine. These quantum advantages persist even when classical machines are granted unlimited time or if BPP=BQP, and rely only on the correctness of quantum mechanics. Together, our results establish machine learning on classical data as a broad and natural domain of quantum advantage and a fundamental test of quantum mechanics at the complexity frontier.
View details
Abstain and Validate: A Dual-LLM Policy for Reducing Noise in Agentic Program Repair
José Cambronero
Michele Tufano
Renyao Wei
Grant Uy
Chin-Jung Liu
Shiying Pan
Satish Chandra
ICSE-SEIP '26: Proceedings of the IEEE/ACM 48th International Conference on Software Engineering: Software Engineering in Practice, ACM (Association for Computing Machinery), New York, NY, USA (2026), 118 – 129
Preview abstract
Agentic automated program repair (APR) is increasingly being used to tackle complex, repository-level bugs in both academic and industry settings. However, changes made by the agent still need to be reviewed by a human before committing them to the codebase to ensure the change correctly fixes the bug. Showing unlikely patches to developers can lead to substantial noise, wasting valuable developer time and decreasing trust in all automated code changes. We introduce two complementary LLM-based policies to reduce such noise: bug abstention and patch critic policies. Bug abstention allows us to exclude bugs that the agentic APR system is unlikely to fix and the patch critic rejects patches that are unlikely to be a good fix for the given bug. We evaluate both policies on three sets of bugs collected from Google’s codebase and agent trajectories generated by an agentic APR system in use at Google. We introduce filtered success rates, where we compute performance after removing trajectories rejected by our policies. On a set of 174 human-reported bugs, abstention and critic can raise filtered success rates up to 13 percentage points and 15 percentage points, respectively, and up to 29 percentage points in combination. For null pointer exceptions and sanitizer-reported bugs with machine-generated bug reports or known reproduction tests, we show the same critic policy can be used to improve average single-sample success rates.
View details
Improving Low-Vision Chart Accessibility via On-Cursor Visual Context
Yotam Sechayk
Hennes Rave
Max Radler
Mark Colley
Ariel Shamir
Takeo Igarashi
Proceedings of the 2026 CHI Conference on Human Factors in Computing Systems (CHI 26)
Preview abstract
Despite widespread use, charts remain largely inaccessible for Low-Vision Individuals (LVI). Reading charts requires viewing data points within a global context, which is difficult for LVI who may rely on magnification or experience a partial field of vision. We aim to improve exploration by providing visual access to critical context. To inform this, we conducted a formative study with five LVI. We identified four fundamental contextual elements common across chart types: axes, legend, grid lines, and the overview. We propose two pointer-based interaction methods to provide this context: Dynamic Context, a novel focus+context interaction, and Mini-map, which adapts overview+detail principles for LVI. In a study with N=22 LVI, we compared both methods and evaluated their integration to current tools. Our results show that Dynamic Context had significant positive impact on access, usability, and effort reduction; however, worsened visual load. Mini-map strengthened spatial understanding, but was less preferred for this task. We offer design insights to guide the development of future systems that support LVI with visual context while balancing visual load.
View details
Preview abstract
The management of a hybrid workforce comprising human and autonomous computational agents may be challenged by the use of separate systems for human capital and software assets, which can create a governance gap. A system can provide a unified framework for managing a hybrid workforce. For example, the system may utilize a labor service mesh to analyze and route tasks to either a human intent tier or an agentic execution tier. A potential principle of the system is structural symmetry, where computational agents can be assigned digital identities and managed through a lifecycle process that may parallel human resource functions, such as onboarding, performance evaluation, and structured offboarding. This integrated approach can facilitate a unified system of record and governance model for an organization's intelligence capacity.
View details
Taming the Variants Multi-Architecture Continuous Testing at Google
Chandrakanth Chittappa
Ali Esmaeeli
Laura Macaddino
Sam Manfreda
David Margolin
Dharma Naidu
Sabuj Pattanayek
Sachin Sable
Ruslan Sakevych
Dushyant Acharya
Adrian Berding
Kevin Crossan
Wolff Dobson
Abhay Singh
19th IEEE International Conference on Software Testing, Verification and Validation (ICST) 2026, Daejeon, Republic of Korea, IEEE
Preview abstract
Enterprises are increasingly adopting multiple general-purpose computer architectures in the data center. This leads to new testing challenges as it creates demand to qualify the software for the additional architectures. Naively double-testing all software for both architectures is costly and unnecessary. Further, reconfiguring CI/CD to take advantage of the new architecture can be non-trivial at scale. This paper introduces CI/CD variants and an optimized testing cycle to solve these twin challenges. We empirically evaluate our solution's impact on human and machine expenses using 44k projects at Google on real production data. First, we estimate saving ~25% of machine expenses at the negligible cost of a few delayed breakage detections per day. Second, we estimate a 90+% reduction in human cost for migrating the configuration. All features described in this paper are now Generally Available at Google and we report this as an empirical case study in scaling CI/CD to new architectures.
View details
