Publications
Papers by DAPLab members, newest first. Lab members are in bold.
2026
- FlowCheck: Helping End-users Specify and Verify Intent in Vibe-Coded Web Apps
- Passes Alone, Fails Together: Benchmarking Semantic Coordination in Parallel LLM-Agent Development
- WebUIProof: Benchmarking WebUI Code Generators with UI-Agent Execution Harness
- StateFork: Branchable Infrastructure for Agent Exploration
- Environment Steering: Using Data Flow Control to Improve Agent Utility and Safety
- SANA: What Matters for QA Agents over Massive Data Lakes?
- SCOUT: Synergizing Reasoning and Tool-Use for Computer-Use Safety
- Receiver-Conditioned Latent Communication gives 94% CacheBack
- SimTrace: Grounded Multimodal User Trajectories Generation for Online User Modeling
- Prove2Me: An Open Collaborative Platform for Scaling Math Formalization
- One Success Isn't Reliability: Thinkingbox, a Sandbox and Benchmark for Agents in Stateful Business Workflows
- The Next Challenge for Agentic Cybersecurity: A Realistic, Contamination-Free Reverse Engineering Benchmark
- OpenForgeRL: Train Harness-native Agents in Any Environment
- TRIM: Reducing AI-Generated CodeSlop via Agent Trajectory Minimization
- fog: Expressing Motion and Emotion through Function Composition of AI-Generated Code
- PersonaJudge: Simulating Individual Human Preference Judgments with Evaluator-Specific Demonstration Data
- From Poisoned to Aware: Fostering Backdoor Self-Awareness in LLMs
- LAKEQA: An Exploratory QA Benchmark over a Million-Scale Data Lake
- Outrunning LLM Cutoffs: A Live Kernel Crash Resolution Benchmark for All
- Latent Cache Flow: Model-to-Model Communication Without Text
- BranchBench: An Extensible Benchmark for Agentic Database Branching
- Behavior Cloning is Not All You Need: The Optimality of On-Policy Distillation for Noisy Expert Feedback
- OSWorld2.0: Benchmarking Computer Use Agents on Long-Horizon Real-World Tasks
- Invisible Impact of Empathy on Behavioral Change: Isolating the Effect of Empathy in Long-term Physical Activity Coaching Chatbot Interactions
- GIF: Locally Sound Geometric Information Flow Control for LLMs
- Skills for the Future Software Profession: Beyond Agentic AI!
- MortarBench: Evaluating Mortgage Loan Origination Agents
- Detect, Remask, Repair: Diffusion Editing for Faithful Summarization of Evolving Contexts
- Fixed Universal Transformer
- VISTA: A Versatile Interactive User Simulation Toolkit for Agent Evaluation
- Data Flow Control: Data Safety Policies for AI Agents
- Orchard: An Open-Source Agentic Modeling Framework
- AutoRPA: Efficient GUI Automation through LLM-Driven Code Synthesis from Interactions
- Detecting Privilege Escalation in Polyglot Microservices via Agentic Program Analysis
- PQR: A Framework to Generate Diverse and Realistic User Queries that Elicit QA Agent Failures
- SemaTune: Semantic-Aware Online OS Tuning with Large Language Models
- Orchard: An Open-Source Agentic Modeling Framework
- Conversational Customization of Productivity Systems: A Design Probe of Malleable AI Interfaces
- ScarfBench: A Benchmark for Cross-Framework Application Migration in Enterprise Java
- Panprediction: optimal predictions for any downstream task and loss
- Prior makes it possible: from sublinear graph algorithms to LLM test-time methods
- Your Compiler is Backdooring Your Model: Understanding and Exploiting Compilation Inconsistency Vulnerabilities in Deep Learning Compilers
- Estimating Tail Risks in Language Model Output Distributions
- BranchBench: Aligning Database Branching with Agentic Demands
- MIND: Empowering Mental Health Clinicians with Multimodal Data Insights through a Narrative Dashboard
- More than Decision Support: Exploring Patients' Longitudinal Usage of Large Language Models in Real-World Healthcare-Seeking Journeys
- VineLM: Trie-Based Fine-Grained Control for Agentic Workflows
- SYN-DIGITS: A Synthetic Control Framework for Calibrated Digital Twin Simulation
- AgentOpt v0.1 Technical Report: Client-Side Optimization for LLM-Based Agents
- Quantifying Trust: Financial Risk Management for Trustworthy AI Agents
- Agentic Data Environments
- Coherence Collapse: Diagnosing Why Code Agents Fail After Reaching the Right Code
- Human-Data Interaction, Exploration, and Visualization in the AI Era: Challenges and Opportunities
- Group-realizable multi-group learning by minimizing empirical risk
- Trustworthy AI Software Engineers
- Please Don't Kill My Vibe: Empowering Agents with Data Flow Control
- AgentWorld: Benchmarking Long-Horizon Collaboration of Multi-agent LLMs
2025
- Multi-Agent Markov Entanglement
- LLM Agents for Always-On Operating System Tuning
- Data Mixture Optimization: A Multi-Fidelity Multi-Scale Bayesian Framework
- Agents for Web Testing: A Case Study in the Wild
- Q-learning with Posterior Sampling
- Touch in the Wild: Learning Fine-Grained Manipulation with a Portable Visuo-Tactile Gripper
- RAISE: Reliable Agent Improvement via Simulated Experience
- LLM Generated Persona is a Promise with a Catch
- Tail-Optimized Caching for LLM Inference
- Fast attention mechanisms: a tale of parallelism
- A Decade of Systems for Human Data Interaction
- SAGE: A Top-Down Bottom-Up Knowledge-Grounded User Simulator for Multi-turn AGent Evaluation
- Cortex: Workflow-Aware Resource Pooling and Scheduling for Agentic Serving
- Toward Systems Foundations for Agentic Exploration
- Set It and Forget It: Zero-Mod ML Magic for Linux Tuning
- Dyna-Mind: Learning to Simulate from Experience for Better AI Agents
- Do Spammers Dream of Electric Sheep? Characterizing the Prevalence of LLM-Generated Malicious Emails
- PickleBall: Secure Deserialization of Pickle-based Machine Learning Models
- The Anatomy of a Personal Health Agent
- Suna: Scalable Causal Confounder Discovery over Relational Data
- EditLord: Learning Code Transformation Rules for Code Editing
- Learning to Rewrite: Generalized LLM-Generated Text Detection
- Performance of LLMs on Stochastic Modeling Operations Research Problems: From Theory to Practice.
- Prompt Editor: A Taxonomy-driven System for Guided LLM Prompt Development in Enterprise Settings
- Towards a Framework for Optimizing Hierarchical Text Segmentation using LLMs
- Position Paper: A System-Centric Approach is Necessary for AI Agents
- Twin-2K-500: A dataset for building digital twins of over 2,000 people based on their answers to over 500 questions
- Diversity Helps Jailbreak Large Language Models
- CrashFixer: A Crash Resolution Agent for the Linux Kernel
- FeedQUAC: Quick Unobtrusive Agent-Generated Commentary
- Steering Semantic Data Processing With DocWrangler
- Throughput-Optimal Scheduling Algorithms for LLM Inference and AI Agents
- AgentDynEx: Nudging the Mechanics and Dynamics of Multi-Agent Simulations
- DocETL: Agentic Query Rewriting and Evaluation for Complex Document Processing
- Simulating Cooperative Prosocial Behavior with Multi-Agent LLMs
- Program Synthesis Dialog Agents for Interactive Decision-Making
- How Well do LLMs Compress their Own Chain-of-Thought? A Token Complexity Approach
- ExACT: Teaching AI Agents to Explore with Reflective-MCTS and Exploratory Learning
- AnimationAgents: A Multi-Modal Team of Agents for Generating, Debugging, and Human Editing of Animation Code
- ACE: A LLM Agent-based Negotiation Coaching System
- Fast Userspace Networking for the Rest of Us
- DynEx: Agentic Assistance to Bridge Design and Code
- DietGlance: dietary monitoring and personalized analysis at a glance with knowledge-empowered AI assistant
2024
Other
- Law-Abiding Agents: Demonstrating Safe Tax Preparation with Data Flow Control