arXiv subjects

George Lee

Publications and source records attributed to George Lee.

13 recordsLinked to original sources

arXiv2026

PLE: Physical-Engineering's Last Exam

Frontier language models now score above 90% on popular knowledge benchmarks and are increasingly deployed as agents that carry out multi-hour professional work. Yet no existing benchmark tests whether they can perform the real-world, economically valuable work of a professional physical engineer, such as modelling physics and legal code requirements across a multi-format project corpus, optimising material and installation costs, and issuing 2D/3D engineering design packages that a licensed reviewer would sign off on. Existing engineering benchmarks focus on textbook question answering, isolated tool calls or self-contained toy design problems. As a result, most of these benchmarks are already saturated by leading agents and solving them is far away from delivering real engineering design. We introduce PLE, a benchmark of 85 long-horizon, agentic tasks in professional physical engineering consisting of electrical, civil, structural and mechanical engineering, where solving a task means completing a realistic, billable engineering deliverable. PLE combines two complementary curation paths. Expert-Authored tasks are created end to end by practising engineers from a professional data provider's expert pool. Co-Authored tasks start from problem statements written by practising engineers; an adversarial Creator-Solver agent pipeline constructs each task's realistic project environment, executable grader and reference deliverables before sending them to expert engineers for validation and difficulty intensification. We evaluate a variety of frontier LLMs under seven publicly available agent harnesses. Even the strongest configuration reaches a mean score of only 50.6%, while spending 27M tokens and 63 minutes per task on average, demonstrating the substantial gap between current agents and autonomous physical engineering design.

cs.AI↗

arXiv2026

RubricEM: Meta-RL with Rubric-guided Policy Decomposition beyond Verifiable Rewards

Deep research agents plan, search, evaluate evidence, and synthesize long-form reports. Training these agents with reinforcement learning (RL) is challenging because they operate over long horizons and their outputs lack exact answer verifiers. To this end, we introduce RubricEM, a unified RL framework for deep research that combines stagewise credit assignment with reflection-based meta-RL for experience reuse. Rubrics unify the framework by providing a shared interface for policy execution, judge feedback, and experience reuse. RubricEM first instills a rubric-guided scaffold through supervised fine-tuning, teaching the agent to plan with rubrics and organize its trajectory into semantic stages. Building on this structural prior, we optimize the task policy with Stage-Structured GRPO, using stagewise judge feedback for finer-grained credit assignment. In parallel, we asynchronously train a shared-backbone reflection meta-policy (with meta-RL) to distill completed trajectories into reusable guidance, contributing through both shared-parameter updates and textual memory. The resulting RubricEM-8B model achieves strong performance across four long-form research benchmarks, outperforming existing open models of larger sizes and approaching proprietary deep-research systems. Beyond final performance, we perform thorough controlled analyses to understand the key ingredients of RubricEM.

cs.CL↗

arXiv2026

PI-Hunter: Automated Red-Teaming for Exposing and Localizing Prompt Injections

Large Language Models (LLMs) are rapidly evolving into agentic systems that interact with external tools and environments, introducing new security risks such as indirect prompt injection attacks through untrusted external sources. Existing defenses mainly focus on blocking malicious content at inference time, and current red-teaming methods primarily optimize attack success. As a result, developers have limited visibility into how latent prompt injections emerge and propagate through agents. We propose PI-Hunter, an automated agentic auditing framework for proactive vulnerability exposure in LLM agents. PI-Hunter constructs realistic source-aware test cases and iteratively evolves them through feedback-driven exploration to induce agents to retrieve and reveal latent malicious instructions embedded within external environments. Extensive experiments across multiple benchmarks, agent architectures, attacks, and defenses demonstrate that PI-Hunter substantially improves vulnerability exposure and attack-surface coverage over strong automated red-teaming baselines, while remaining effective under existing prompt injection defenses.

cs.CR↗

arXiv2026

SkillOS: Learning Skill Curation for Self-Evolving Agents

LLM-based agents are increasingly deployed to handle streaming tasks, yet they often remain one-off problem solvers that fail to learn from past interactions. Reusable skills distilled from experience provide a natural substrate for self-evolution, where high-quality skill curation serves as the key bottleneck. Existing approaches either rely on manual skill curation, prescribe heuristic skill operations, or train for short-horizon skill operations. However, they still struggle to learn complex long-term curation policies from indirect and delayed feedback. To tackle this challenge, we propose SkillOS, an experience-driven RL training recipe for learning skill curation in self-evolving agents. SkillOS pairs a frozen agent executor that retrieves and applies skills with a trainable skill curator that updates an external SkillRepo from accumulated experience. To provide learning signals for curation, we design composite rewards and train on grouped task streams based on skill-relevant task dependencies, where earlier trajectories update the SkillRepo, and later related tasks evaluate these updates. Across multi-turn agentic tasks and single-turn reasoning tasks, SkillOS consistently outperforms memory-free and strong memory-based baselines in both effectiveness and efficiency, with the learned skill curator generalizing across different executor backbones and task domains. Further analyses show that the learned curator produces more targeted skill use, while the skills in SkillRepo evolve into more richly structured Markdown files that encode higher-level meta-skills over time.

cs.AI↗

arXiv2026

ReasoningBank: Scaling Agent Self-Evolving with Reasoning Memory

With the growing adoption of large language model agents in persistent real-world roles, they naturally encounter continuous streams of tasks. A key limitation, however, is their failure to learn from the accumulated interaction history, forcing them to discard valuable insights and repeat past errors. We propose ReasoningBank, a novel memory framework that distills generalizable reasoning strategies from an agent's self-judged successful and failed experiences. At test time, an agent retrieves relevant memories from ReasoningBank to inform its interaction and then integrates new learnings back, enabling it to become more capable over time. Building on this powerful experience learner, we further introduce memory-aware test-time scaling (MaTTS), which accelerates and diversifies this learning process by scaling up the agent's interaction experience. By allocating more compute to each task, the agent generates abundant, diverse experiences that provide rich contrastive signals for synthesizing higher-quality memory. The better memory in turn guides more effective scaling, establishing a powerful synergy between memory and test-time scaling. Across web browsing and software engineering benchmarks, ReasoningBank consistently outperforms existing memory mechanisms that store raw trajectories or only successful task routines, improving both effectiveness and efficiency; MaTTS further amplifies these gains. These findings establish memory-driven experience scaling as a new scaling dimension, enabling agents to self-evolve with emergent behaviors naturally arise. Our code can be found at https://github.com/google-research/reasoning-bank.

cs.AI↗

arXiv2025

In Prospect and Retrospect: Reflective Memory Management for Long-term Personalized Dialogue Agents

Large Language Models (LLMs) have made significant progress in open-ended dialogue, yet their inability to retain and retrieve relevant information from long-term interactions limits their effectiveness in applications requiring sustained personalization. External memory mechanisms have been proposed to address this limitation, enabling LLMs to maintain conversational continuity. However, existing approaches struggle with two key challenges. First, rigid memory granularity fails to capture the natural semantic structure of conversations, leading to fragmented and incomplete representations. Second, fixed retrieval mechanisms cannot adapt to diverse dialogue contexts and user interaction patterns. In this work, we propose Reflective Memory Management (RMM), a novel mechanism for long-term dialogue agents, integrating forward- and backward-looking reflections: (1) Prospective Reflection, which dynamically summarizes interactions across granularities-utterances, turns, and sessions-into a personalized memory bank for effective future retrieval, and (2) Retrospective Reflection, which iteratively refines the retrieval in an online reinforcement learning (RL) manner based on LLMs' cited evidence. Experiments show that RMM demonstrates consistent improvement across various metrics and benchmarks. For example, RMM shows more than 10% accuracy improvement over the baseline without memory management on the LongMemEval dataset.

cs.CL↗

arXiv2025

Deep Researcher with Test-Time Diffusion

Deep research agents, powered by Large Language Models (LLMs), are rapidly advancing; yet, their performance often plateaus when generating complex, long-form research reports using generic test-time scaling algorithms. Drawing inspiration from the iterative nature of human research, which involves cycles of searching, reasoning, and revision, we propose the Test-Time Diffusion Deep Researcher (TTD-DR). This novel framework conceptualizes research report generation as a diffusion process. TTD-DR initiates this process with a preliminary draft, an updatable skeleton that serves as an evolving foundation to guide the research direction. The draft is then iteratively refined through a "denoising" process, which is dynamically informed by a retrieval mechanism that incorporates external information at each step. The core process is further enhanced by a self-evolutionary algorithm applied to each component of the agentic workflow, ensuring the generation of high-quality context for the diffusion process. This draft-centric design makes the report writing process more timely and coherent while reducing information loss during the iterative search process. We demonstrate that our TTD-DR achieves state-of-the-art results on a wide array of benchmarks that require intensive search and multi-hop reasoning, significantly outperforming existing deep research agents.

cs.CL↗

arXiv2022

Constructing the Oseledets decomposition with subspace growth estimates

The semi-invertible version of Oseledets' multiplicative ergodic theorem providing a decomposition of the underlying state space of a random linear dynamical system into fast and slow spaces is deduced for a strongly measurable cocycle on a separable Banach space. This work represents a significantly simplified means of obtaining the result, using measurable growth estimates on subspaces for linear operators combined with a modified version of Kingman's subadditive ergodic theorem.

math.DS↗

arXiv2020

Adaptive Charging Networks: A Framework for Smart Electric Vehicle Charging

We describe the architecture and algorithms of the Adaptive Charging Network (ACN), which was first deployed on the Caltech campus in early 2016 and is currently operating at over 100 other sites in the United States. The architecture enables real-time monitoring and control and supports electric vehicle (EV) charging at scale. The ACN adopts a flexible Adaptive Scheduling Algorithm based on convex optimization and model predictive control and allows for significant over-subscription of electrical infrastructure. We describe some of the practical challenges in real-world charging systems, including unbalanced three-phase infrastructure, non-ideal battery charging behavior, and quantized control signals. We demonstrate how the Adaptive Scheduling Algorithm handles these challenges, and compare its performance against baseline algorithms from the deadline scheduling literature using real workloads recorded from the Caltech ACN and accurate system models. We find that in these realistic settings, our scheduling algorithm can improve operator profit by 3.4 times over uncontrolled charging and consistently outperforms baseline algorithms when delivering energy in highly congested systems.

eess.SY↗

arXiv2012

The Unified Logging Infrastructure for Data Analytics at Twitter

In recent years, there has been a substantial amount of work on large-scale data analytics using Hadoop-based platforms running on large clusters of commodity machines. A less-explored topic is how those data, dominated by application logs, are collected and structured to begin with. In this paper, we present Twitter's production logging infrastructure and its evolution from application-specific logging to a unified "client events" log format, where messages are captured in common, well-formatted, flexible Thrift messages. Since most analytics tasks consider the user session as the basic unit of analysis, we pre-materialize "session sequences", which are compact summaries that can answer a large class of common queries quickly. The development of this infrastructure has streamlined log collection and data analysis, thereby improving our ability to rapidly experiment and iterate on various aspects of the service.

cs.DB↗

arXiv2008

Double bubbles in $S^3$ and $H^3$

We prove the double bubble conjecture in the three-sphere $S^3$ and hyperbolic three-space $H^3$ in the cases where we can apply Hutchings theory: 1) in $S^3$, each enclosed volume and the complement occupy at least 10% of the volume of $S^3$; 2) in $H^3$, the smaller volume is at least 85% that of the larger. A balancing argument and asymptotic analysis reduce the problem in $S^3$ and $H^3$ to some computer checking. The computer analysis has been designed and fully implemented for both spaces.

math.DG↗

arXiv2003

The Double Bubble Problem on the Flat Two-Torus

We characterize the perimeter-minimizing double bubbles on all flat two-tori and, as corollaries, on the flat infinite cylinder and the flat infinite strip with free boundary. Specifically, we show that there are five distinct types of minimizers on flat two-tori, depending on the areas to be enclosed.

math.MG↗

arXiv1998

Microsoft TerraServer

The Microsoft TerraServer stores aerial and satellite images of the earth in a SQL Server Database served to the public via the Internet. It is the world's largest atlas, combining five terabytes of image data from the United States Geodetic Survey, Sovinformsputnik, and Encarta Virtual Globe. Internet browsers provide intuitive spatial and gazetteer interfaces to the data. The TerraServer is also an E-Commerce application. Users can buy the right to use the imagery using Microsoft Site Servers managed by the USGS and Aerial Images. This paper describes the TerraServer's design and implementation.

cs.DB↗