AIConfPaper
← Search

Percy Liang

147 accepted papers

2026

Comparing AI Agents to Cybersecurity Professionals in Real-World Penetration Testing

Justin W Lin, Eliot Krzysztof Jones, Donovan Julian Jasper, Ethan Jun-shen Ho, Anna Wu, Arnold Tianyi Yang +7
ICLR 2026poster

We present the first comprehensive evaluation of AI agents against human cybersecurity professionals in a live enterprise environment. We evaluate ten cybersecurity professionals alongside six existing AI agents and ARTEMIS, our new agent scaffold, on a large university network consisting of $\sim$8…

Cited by 0Source ↗code ↗Scholar ↗
2026

Cosmos Policy: Fine-Tuning Video Models for Visuomotor Control and Planning

Moo Jin Kim, Yihuai Gao, Tsung-Yi Lin, Yen-Chen Lin, Yunhao Ge, Grace Lam +5
ICLR 2026poster

Recent video generation models demonstrate remarkable ability to capture complex physical interactions and scene evolution over time. To leverage their spatiotemporal priors, robotics works have adapted video models for policy learning but introduce complexity by requiring multiple stages of post-tr…

Cited by 0Source ↗code ↗Scholar ↗
2026

Holistic Agent Leaderboard: The Missing Infrastructure for AI Agent Evaluation

Sayash Kapoor, Benedikt Stroebl, Peter Kirgis, Nitya Nadgir, Zachary S Siegel, Boyi Wei +25
ICLR 2026poster

AI agents have been developed for complex real-world tasks from coding to customer service. But AI agent evaluations suffer from many challenges that undermine our understanding of how well agents really work (Figure 1). We introduce the Holistic Agent Leaderboard (HAL) to address these challenges.…

Cited by 0Source ↗code ↗Scholar ↗
2026

MLE-Smith: Scaling MLE Tasks with Automated Multi-agent Pipeline

Rushi Qiang, Yuchen Zhuang, Anikait Singh, Percy Liang, Chao Zhang, Sherry Yang +1
ICLR 2026poster

While Language Models (LMs) have made significant progress in automating machine learning engineering (MLE), the acquisition of high-quality MLE training data is significantly constrained. Current MLE benchmarks suffer from low scalability and limited applicability because they rely on static, manua…

Cited by 0Source ↗Scholar ↗
2026

VLAW: Iterative Co-Improvement of Vision-Language-Action Policy and World Model

Yanjiang Guo, Tony Lee, Lucy Xiaoyang Shi, Jianyu Chen, Percy Liang, Chelsea Finn
ICML 2026poster

The goal of this paper is to improve the performance and reliability of vision-language-action (VLA) models through iterative online interaction. Since collecting policy rollouts in the real world is expensive, we investigate whether a learned simulator—specifically, an action-conditioned video gene…

Cited by 0Source ↗Scholar ↗
2026

WorldGym: World Model as An Environment for Policy Evaluation

Julian Hector Quevedo, Ansh Kumar Sharma, Yixiang Sun, Varad Suryavanshi, Percy Liang, Sherry Yang
ICLR 2026poster

Evaluating robot control policies is difficult: real-world testing is costly, and handcrafted simulators require manual effort to improve in realism and generality. We propose a world-model-based policy evaluation environment (WorldGym), an autoregressive, action-conditioned video generation model w…

Cited by 0Source ↗Scholar ↗
2025

AIR-BENCH 2024: A Safety Benchmark based on Regulation and Policies Specified Risk Categories

Yi Zeng, Yu Yang, Andy Zhou, Jeffrey Ziwei Tan, Yuheng Tu, Yifan Mai +6
ICLR 2025spotlight

Foundation models (FMs) provide societal benefits but also amplify risks. Governments, companies, and researchers have proposed regulatory frameworks, acceptable use policies, and safety benchmarks in response. However, existing public benchmarks often define safety categories based on previous lite…

Cited by 0Source ↗PDF ↗Scholar ↗
2025

Auditing Prompt Caching in Language Model APIs

Chenchen Gu, Xiang Lisa Li, Rohith Kuditipudi, Percy Liang, Tatsunori Hashimoto
ICML 2025poster

Prompt caching in large language models (LLMs) results in data-dependent timing variations: cached prompts are processed faster than non-cached prompts. These timing differences introduce the risk of side-channel timing attacks. For example, if the cache is shared across users, an attacker could ide…

Cited by 0Source ↗PDF ↗code ↗Scholar ↗
2025

Audits Under Resource, Data, and Access Constraints: Scaling Laws For Less Discriminatory Alternatives

Sarah H. Cen, Salil Goyal, Zaynah Javed, Ananya Karthik, Percy Liang, Daniel E. Ho
NeurIPS 2025poster

AI audits play a critical role in AI accountability and safety. They are particularly salient in anti-discrimination law. Several areas of anti-discrimination law implicate what is known as the "less discriminatory alternative" (LDA) requirement, under which a protocol is defensible if no less discr…

Cited by 0Source ↗Scholar ↗
2025

AutoBencher: Towards Declarative Benchmark Construction

Xiang Lisa Li, Farzaan Kaiyom, Evan Zheran Liu, Yifan Mai, Percy Liang, Tatsunori Hashimoto
ICLR 2025poster

We present AutoBencher, a declarative framework for automatic benchmark construction, and use it to scalably discover novel insights and vulnerabilities of existing language models. Concretely, given a few desiderata of benchmarks (e.g., question difficulty, topic salience), we operationalize each d…

Cited by 0Source ↗PDF ↗Scholar ↗
2025

BioDiscoveryAgent: An AI Agent for Designing Genetic Perturbation Experiments

Yusuf H Roohani, Andrew H. Lee, Qian Huang, Jian Vora, Zachary Steinhart, Kexin Huang +3
ICLR 2025poster

Agents based on large language models have shown great potential in accelerating scientific discovery by leveraging their rich background knowledge and reasoning capabilities. In this paper, we introduce BioDiscoveryAgent, an agent that designs new experiments, reasons about their outcomes, and effi…

Cited by 16Source ↗PDF ↗code ↗Scholar ↗
2025

Blackbox Model Provenance via Palimpsestic Membership Inference

Rohith Kuditipudi, Jing Huang, Sally Zhu, Diyi Yang, Christopher Potts, Percy Liang
NeurIPS 2025spotlight

Suppose Alice trains an open-weight language model and Bob uses a blackbox derivative of Alice’s model to produce text. Can Alice prove that Bob is using her model, either by querying Bob’s derivative model (query setting) or from the text alone ( observational setting)? We formulate this question a…

Cited by 0Source ↗Scholar ↗
2025

BountyBench: Dollar Impact of AI Agent Attackers and Defenders on Real-World Cybersecurity Systems

Andy K Zhang, Joey Ji, Celeste Menders, Riya Dulepet, Thomas Qin, Ron Yifeng Wang +28
NeurIPS 2025poster

AI agents have the potential to significantly alter the cybersecurity landscape. Here, we introduce the first framework to capture offensive and defensive cyber-capabilities in evolving real-world systems. Instantiating this framework with BountyBench, we set up 25 systems with complex, real-world c…

Cited by 0Source ↗Scholar ↗
2025

Cybench: A Framework for Evaluating Cybersecurity Capabilities and Risks of Language Models

Andy K Zhang, Neil Perry, Riya Dulepet, Joey Ji, Celeste Menders, Justin W Lin +19
ICLR 2025oral

Language Model (LM) agents for cybersecurity that are capable of autonomously identifying vulnerabilities and executing exploits have potential to cause real-world impact. Policymakers, model providers, and researchers in the AI and cybersecurity communities are interested in quantifying the capabil…

Cited by 33Source ↗PDF ↗Scholar ↗
2025

Eliciting Language Model Behaviors with Investigator Agents

Xiang Lisa Li, Neil Chowdhury, Daniel D. Johnson, Tatsunori Hashimoto, Percy Liang, Sarah Schwettmann +1
ICML 2025poster

Language models exhibit complex, diverse behaviors when prompted with free-form text, making it hard to characterize the space of possible outputs. We study the problem of behavioral elicitation, where the goal is to search for prompts that induce specific target behaviors (e.g., hallucinations, har…

Cited by 3Source ↗PDF ↗Scholar ↗
2025

Establishing Best Practices in Building Rigorous Agentic Benchmarks

Yuxuan Zhu, Tengjun Jin, Yada Pruksachatkun, Andy K Zhang, Shu Liu, Sasha Cui +20
NeurIPS 2025poster

Benchmarks are essential for quantitatively tracking progress in AI. As AI agents become increasingly capable, researchers and practitioners have introduced agentic benchmarks to evaluate agents on complex, real-world tasks. These benchmarks typically measure agent capabilities by evaluating task ou…

Cited by 0Source ↗Scholar ↗
2025

Language Models May Verbatim Complete Text They Were Not Explicitly Trained On

Ken Liu, Christopher A. Choquette-Choo, Matthew Jagielski, Peter Kairouz, Sanmi Koyejo, Percy Liang +1
ICML 2025spotlight

An important question today is whether a given text was used to train a large language model (LLM). A completion test is often employed: check if the LLM completes a sufficiently complex text. This, however, requires a ground-truth definition of membership; most commonly, it is defined as a member b…

Cited by 0Source ↗PDF ↗Scholar ↗
2025

LawInstruct: A Resource for Studying Language Model Adaptation to the Legal Domain

Joel Niklaus, Lucia Zheng, Arya D. McCarthy, Christopher Hahn, Brian M Rosen, Peter Henderson +4
NAACL 2025findings

Instruction tuning is an important step in making language models useful for direct user interaction. However, the legal domain is underrepresented in typical instruction datasets (e.g., only 10 out of 1600+ tasks in Super-NaturalInstructions). To study whether instruction tuning on legal datasets i…

Cited by 1Source ↗PDF ↗code ↗Scholar ↗
2025

MLE-Dojo: Interactive Environments for Empowering LLM Agents in Machine Learning Engineering

Rushi Qiang, Yuchen Zhuang, Yinghao Li, Dingu Sagar V K, Rongzhi Zhang, ChangHao Li +5
NeurIPS 2025poster

We introduce MLE-Dojo, a Gym-style framework for systematically reinforcement learning, evaluating, and improving autonomous large language model (LLM) agents in iterative machine learning engineering (MLE) workflows. Unlike existing benchmarks that primarily rely on static datasets or single-attemp…

Cited by 0Source ↗code ↗Scholar ↗
2025

Machine Unlearning Doesn't Do What You Think: Lessons for Generative AI Policy and Research

A. Feder Cooper, Christopher A. Choquette-Choo, Miranda Bogen, Kevin Klyman, Matthew Jagielski, Katja Filippova +31
NeurIPS 2025oral

"Machine unlearning" is a popular proposed solution for mitigating the existence of content in an AI model that is problematic for legal or moral reasons, including privacy, copyright, safety, and more. For example, unlearning is often invoked as a solution for removing the effects of specific infor…

Cited by 0Source ↗Scholar ↗
2025

Position: In-House Evaluation Is Not Enough. Towards Robust Third-Party Evaluation and Flaw Disclosure for General-Purpose AI

Shayne Longpre, Kevin Klyman, Ruth Elisabeth Appel, Sayash Kapoor, Rishi Bommasani, Michelle Sahar +28
ICML 2025spotlight

The widespread deployment of general-purpose AI (GPAI) systems introduces significant new risks. Yet the infrastructure, practices, and norms for reporting flaws in GPAI systems remain seriously underdeveloped, lagging far behind more established fields like software security. Based on a collaborati…

Cited by 0Source ↗PDF ↗Scholar ↗
2025

Position: Language model developers should report train-test overlap

Andy K Zhang, Kevin Klyman, Yifan Mai, Yoav Levine, Yian Zhang, Rishi Bommasani +1
ICML 2025spotlight

Language models are extensively evaluated, but correctly interpreting evaluation results requires knowledge of train-test overlap, which refers to the extent to which the language model is trained on the very data it is being tested on. The public currently lacks adequate information about train-tes…

Cited by 6Source ↗PDF ↗Scholar ↗
2025

Reliable and Efficient Amortized Model-based Evaluation

Sang T. Truong, Yuheng Tu, Percy Liang, Bo Li, Sanmi Koyejo
ICML 2025poster

Comprehensive evaluations of language models (LM) during both development and deployment phases are necessary because these models are thought to possess numerous capabilities as well as safety risks. The average score across a wide range of benchmarks provides a signal that helps guide the use of t…

Cited by 2Source ↗PDF ↗Scholar ↗
2025

RoboArena: Distributed Real-World Evaluation of Generalist Robot Policies

Pranav Atreya, Karl Pertsch, Tony Lee, Moo Jin Kim, Arhan Jain, Artur Kuramshin +20
CoRL 2025oral

Comprehensive, unbiased, and comparable evaluation of modern generalist policies is uniquely challenging: existing approaches for robot benchmarking typically rely on heavy standardization, either by specifying fixed evaluation tasks and environments, or by hosting centralized "robot challenges", an…

Cited by 0Source ↗Scholar ↗
2025

Understanding Warmup-Stable-Decay Learning Rates: A River Valley Loss Landscape View

Kaiyue Wen, Zhiyuan Li, Jason S. Wang, David Leo Wright Hall, Percy Liang, Tengyu Ma
ICLR 2025poster

Training language models currently requires pre-determining a fixed compute budget because the typical cosine learning rate schedule depends on the total number of steps. In contrast, the Warmup-Stable-Decay (WSD) schedule uses a constant learning rate to produce a main branch of iterates that can i…

Cited by 0Source ↗PDF ↗Scholar ↗
2025

s1: Simple test-time scaling

Niklas Muennighoff, Zitong Yang, Weijia Shi, Xiang Lisa Li, Li Fei-Fei, Hannaneh Hajishirzi +4
EMNLP 2025

Test-time scaling is a promising new approach to language modeling that uses extra test-time compute to improve performance. Recently, OpenAI’s o1 model showed this capability but did not publicly share its methodology, leading to many replication efforts. We seek the simplest approach to achieve te

Cited by 0Source ↗PDF ↗code ↗Scholar ↗
2024

Benchmarking and Improving Generator-Validator Consistency of Language Models

Xiang Lisa Li, Vaishnavi Shrivastava, Siyan Li, Tatsunori Hashimoto, Percy Liang
ICLR 2024poster

As of September 2023, ChatGPT correctly answers “what is 7+8” with 15, but when asked “7+8=15, True or False” it responds with “False”. This inconsistency between generating and validating an answer is prevalent in language models (LMs) and erodes trust. In this paper, we propose a framework for mea…

Cited by 29Source ↗PDF ↗Scholar ↗
2024

Embodied Agent Interface: Benchmarking LLMs for Embodied Decision Making

Manling Li, Shiyu Zhao, Qineng Wang, Kangrui Wang, Yu Zhou, Sanjana Srivastava +9
NeurIPS 2024oral

We aim to evaluate Large Language Models (LLMs) for embodied decision making. While a significant body of work has been leveraging LLMs for decision making in embodied environments, we still lack a systematic understanding of their performance because they are usually applied in different domains, f…

Cited by 33Source ↗PDF ↗Scholar ↗
2024

Image2Struct: Benchmarking Structure Extraction for Vision-Language Models

Josselin Somerville Roberts, Tony Lee, Chi Heem Wong, Michihiro Yasunaga, Yifan Mai, Percy Liang
NeurIPS 2024poster

We introduce Image2Struct, a benchmark to evaluate vision-language models (VLMs) on extracting structure from images. Our benchmark 1) captures real-world use cases, 2) is fully automatic and does not require human judgment, and 3) is based on a renewable stream of fresh data. In Image2Struct, VLMs…

Cited by 3Source ↗code ↗Scholar ↗
2024

Large Language Models as Analogical Reasoners

Michihiro Yasunaga, Xinyun Chen, Yujia Li, Panupong Pasupat, Jure Leskovec, Percy Liang +2
ICLR 2024poster

Chain-of-thought (CoT) prompting for language models demonstrates impressive performance across reasoning tasks, but typically needs labeled exemplars of the reasoning process. In this work, we introduce a new prompting approach, analogical prompting, designed to automatically guide the reasoning pr…

Cited by 60Source ↗PDF ↗Scholar ↗
2024

MLAgentBench: Evaluating Language Agents on Machine Learning Experimentation

Qian Huang, Jian Vora, Percy Liang, Jure Leskovec
ICML 2024poster

A central aspect of machine learning research is experimentation, the process of designing and running experiments, analyzing the results, and iterating towards some positive outcome (e.g., improving accuracy). Could agents driven by powerful language models perform machine learning experimentation…

Cited by 61Source ↗PDF ↗code ↗Scholar ↗
2024

MedAlign: A Clinician-Generated Dataset for Instruction Following with Electronic Medical Records

Scott L. Fleming, Alejandro Lozano, William J. Haberkorn, Jenelle A. Jindal, Eduardo Reis, Rahul Thapa +24
AAAI 2024technical

The ability of large language models (LLMs) to follow natural language instructions with human-level fluency suggests many opportunities in healthcare to reduce administrative burden and improve quality of care. However, evaluating LLMs on realistic text generation tasks for healthcare remains chall…

Cited by 67Source ↗PDF ↗Scholar ↗
2024

On the Learnability of Watermarks for Language Models

Chenchen Gu, Xiang Lisa Li, Percy Liang, Tatsunori Hashimoto
ICLR 2024poster

Watermarking of language model outputs enables statistical detection of model-generated text, which can mitigate harms and misuses of language models. Existing watermarking strategies operate by altering the decoder of an existing language model. In this paper, we ask whether language models can dir…

Cited by 44Source ↗PDF ↗code ↗Scholar ↗
2024

OpenVLA: An Open-Source Vision-Language-Action Model

Moo Jin Kim, Karl Pertsch, Siddharth Karamcheti, Ted Xiao, Ashwin Balakrishna, Suraj Nair +11
CoRL 2024poster

Large policies pretrained on a combination of Internet-scale vision-language data and diverse robot demonstrations have the potential to change how we teach robots new skills: rather than training new behaviors from scratch, we can fine-tune such vision-language-action (VLA) models to obtain robust,…

Cited by 437Source ↗Scholar ↗
2024

Position: A Safe Harbor for AI Evaluation and Red Teaming

Shayne Longpre, Sayash Kapoor, Kevin Klyman, Ashwin Ramaswami, Rishi Bommasani, Borhane Blili-Hamelin +17
ICML 2024oral

Independent evaluation and red teaming are critical for identifying the risks posed by generative AI systems. However, the terms of service and enforcement strategies used by prominent AI companies to deter model misuse have disincentives on good faith safety evaluations. This causes some researcher…

Cited by 5Source ↗PDF ↗Scholar ↗
2024

Position: On the Societal Impact of Open Foundation Models

Sayash Kapoor, Rishi Bommasani, Kevin Klyman, Shayne Longpre, Ashwin Ramaswami, Peter Cihon +19
ICML 2024oral

Foundation models are powerful technologies: how they are released publicly directly shapes their societal impact. In this position paper, we focus on *open* foundation models, defined here as those with broadly available model weights (e.g., Llama 3, Stable Diffusion XL). We identify five distincti…

Cited by 4Source ↗PDF ↗Scholar ↗
2024

Prismatic VLMs: Investigating the Design Space of Visually-Conditioned Language Models

Siddharth Karamcheti, Suraj Nair, Ashwin Balakrishna, Percy Liang, Thomas Kollar, Dorsa Sadigh
ICML 2024poster

Visually-conditioned language models (VLMs) have seen growing adoption in applications such as visual dialogue, scene understanding, and robotic task planning; adoption that has fueled a wealth of new models such as LLaVa, InstructBLIP, and PaLI-3. Despite the volume of new releases, key design deci…

Cited by 102Source ↗PDF ↗Scholar ↗
2024

RedPajama: an Open Dataset for Training Large Language Models

Maurice Weber, Daniel Y Fu, Quentin Gregory Anthony, Yonatan Oren, Shane Adams, Anton Alexandrov +13
NeurIPS 2024spotlight

Large language models are increasingly becoming a cornerstone technology in artificial intelligence, the sciences, and society as a whole, yet the optimal strategies for dataset composition and filtering remain largely elusive. Many of the top-performing models lack transparency in their dataset cur…

Cited by 49Source ↗PDF ↗code ↗Scholar ↗
2024

Sophia: A Scalable Stochastic Second-order Optimizer for Language Model Pre-training

Hong Liu, Zhiyuan Li, David Leo Wright Hall, Percy Liang, Tengyu Ma
ICLR 2024poster

Given the massive cost of language model pre-training, a non-trivial improvement of the optimization algorithm would lead to a material reduction on the time and cost of training. Adam and its variants have been state-of-the-art for years, and more sophisticated second-order (Hessian-based) optimize…

Cited by 151Source ↗PDF ↗Scholar ↗
2024

VHELM: A Holistic Evaluation of Vision Language Models

Tony Lee, Haoqin Tu, Chi Heem Wong, Wenhao Zheng, Yiyang Zhou, Yifan Mai +5
NeurIPS 2024poster

Current benchmarks for assessing vision-language models (VLMs) often focus on their perception or problem-solving capabilities and neglect other critical aspects such as fairness, multilinguality, or toxicity. Furthermore, they differ in their evaluation procedures and the scope of the evaluation, m…

Cited by 15Source ↗PDF ↗code ↗Scholar ↗
2024

Vocal Sandbox: Continual Learning and Adaptation for Situated Human-Robot Collaboration

Jennifer Grannen, Siddharth Karamcheti, Suvir Mirchandani, Percy Liang, Dorsa Sadigh
CoRL 2024poster

We introduce Vocal Sandbox, a framework for enabling seamless human-robot collaboration in situated environments. Systems in our framework are characterized by their ability to *adapt and continually learn* at multiple levels of abstraction from diverse teaching modalities such as spoken dialogue, o…

Cited by 0Source ↗code ↗Scholar ↗
2023

AlpacaFarm: A Simulation Framework for Methods that Learn from Human Feedback

Yann Dubois, Xuechen Li, Rohan Taori, Tianyi Zhang, Ishaan Gulrajani, Jimmy Ba +3
NeurIPS 2023spotlight

Large language models (LLMs) such as ChatGPT have seen widespread adoption due to their ability to follow user instructions well. Developing these LLMs involves a complex yet poorly understood workflow requiring training with human feedback. Replicating and understanding this instruction-following p…

Cited by 523Source ↗PDF ↗Scholar ↗
2023

Beyond Positive Scaling: How Negation Impacts Scaling Trends of Language Models

Yuhui Zhang, Michihiro Yasunaga, Zhengping Zhou, Jeff Z. HaoChen, James Zou, Percy Liang +1
ACL 2023findings

Language models have been shown to exhibit positive scaling, where performance improves as models are scaled up in terms of size, compute, or data. In this work, we introduce NeQA, a dataset consisting of questions with negation in which language models do not exhibit straightforward positive scalin…

Cited by 13Source ↗PDF ↗code ↗Scholar ↗
2023

Cheaply Estimating Inference Efficiency Metrics for Autoregressive Transformer Models

Deepak Narayanan, Keshav Santhanam, Peter Henderson, Rishi Bommasani, Tony Lee, Percy Liang
NeurIPS 2023poster

Large language models (LLMs) are highly capable but also computationally expensive. Characterizing the _fundamental tradeoff_ between inference efficiency and model capabilities is thus important, but requires an efficiency metric that is comparable across models from different providers. Unfortuna…

Cited by 7Source ↗PDF ↗code ↗Scholar ↗
2023

CocktailSGD: Fine-tuning Foundation Models over 500Mbps Networks

Jue WANG, Yucheng Lu, Binhang Yuan, Beidi Chen, Percy Liang, Christopher De Sa +2
ICML 2023poster

Distributed training of foundation models, especially large language models (LLMs), is communication-intensive and so has heavily relied on centralized data centers with fast interconnects. Can we train on slow networks and unlock the potential of decentralized infrastructure for foundation models?…

Cited by 49Source ↗PDF ↗Scholar ↗
2023

Contrastive Decoding: Open-ended Text Generation as Optimization

Xiang Lisa Li, Ari Holtzman, Daniel Fried, Percy Liang, Jason Eisner, Tatsunori Hashimoto +2
ACL 2023long

Given a language model (LM), maximum probability is a poor decoding objective for open-ended generation, because it produces short and repetitive text. On the other hand, sampling can often produce incoherent text that drifts from the original topics. We propose contrastive decoding (CD), a reliable…

Cited by 322Source ↗PDF ↗code ↗Scholar ↗
2023

Data Selection for Language Models via Importance Resampling

Sang Michael Xie, Shibani Santurkar, Tengyu Ma, Percy Liang
NeurIPS 2023poster

Selecting a suitable pretraining dataset is crucial for both general-domain (e.g., GPT-3) and domain-specific (e.g., Codex) language models (LMs). We formalize this problem as selecting a subset of a large raw unlabeled dataset to match a desired target distribution given unlabeled target samples. D…

Cited by 183Source ↗PDF ↗code ↗Scholar ↗
2023

Do Question Answering Modeling Improvements Hold Across Benchmarks?

Nelson F. Liu, Tony Lee, Robin Jia, Percy Liang
ACL 2023long

Do question answering (QA) modeling improvements (e.g., choice of architecture and training procedure) hold consistently across the diverse landscape of QA benchmarks? To study this question, we introduce the notion of concurrence—two benchmarks have high concurrence on a set of modeling approaches…

Cited by 9Source ↗PDF ↗code ↗Scholar ↗
2023

DoReMi: Optimizing Data Mixtures Speeds Up Language Model Pretraining

Sang Michael Xie, Hieu Pham, Xuanyi Dong, Nan Du, Hanxiao Liu, Yifeng Lu +4
NeurIPS 2023spotlight

The mixture proportions of pretraining data domains (e.g., Wikipedia, books, web text) greatly affect language model (LM) performance. In this paper, we propose Domain Reweighting with Minimax Optimization (DoReMi), which first trains a small proxy model using group distributionally robust optimizat…

Cited by 163Source ↗PDF ↗code ↗Scholar ↗
2023

Ecosystem-level Analysis of Deployed Machine Learning Reveals Homogeneous Outcomes

Connor Toups, Rishi Bommasani, Kathleen Creel, Sarah H Bana, Dan Jurafsky, Percy Liang
NeurIPS 2023poster

Machine learning is traditionally studied at the model level: researchers measure and improve the accuracy, robustness, bias, efficiency, and other dimensions of specific models. In practice, however, the societal impact of any machine learning model is partially determined by the context into which…

Cited by 11Source ↗PDF ↗Scholar ↗
2023

FlexGen: High-Throughput Generative Inference of Large Language Models with a Single GPU

Ying Sheng, Lianmin Zheng, Binhang Yuan, Zhuohan Li, Max Ryabinin, Beidi Chen +4
ICML 2023oral

The high computational and memory requirements of large language model (LLM) inference make it feasible only with multiple high-end accelerators. Motivated by the emerging demand for latency-insensitive tasks with batched processing, this paper initiates the study of high-throughput LLM inference us…

Cited by 421Source ↗PDF ↗code ↗Scholar ↗
2023

Holistic Evaluation of Text-to-Image Models

Tony Lee, Michihiro Yasunaga, Chenlin Meng, Yifan Mai, Joon Sung Park, Agrim Gupta +12
NeurIPS 2023spotlight

The stunning qualitative improvement of text-to-image models has led to their widespread attention and adoption. However, we lack a comprehensive quantitative understanding of their capabilities and risks. To fill this gap, we introduce a new benchmark, Holistic Evaluation of Text-to-Image Models (H…

Cited by 148Source ↗PDF ↗code ↗Scholar ↗
2023

Is a Caption Worth a Thousand Images? A Study on Representation Learning

Shibani Santurkar, Yann Dubois, Rohan Taori, Percy Liang, Tatsunori Hashimoto
ICLR 2023poster

The development of CLIP [Radford et al., 2021] has sparked a debate on whether adding language supervision can yield vision models with more transferable representations than traditional image-only methods. Our work studies this question through a carefully controlled comparison of two approaches, i…

Cited by 15Source ↗PDF ↗Scholar ↗
2023

Language-Driven Representation Learning for Robotics

Siddharth Karamcheti, Suraj Nair, Annie S Chen, Thomas Kollar, Chelsea Finn, Dorsa Sadigh +1
RSS 2023poster

Recent work in visual representation learning for robotics demonstrates the viability of learning from large video datasets of humans performing everyday tasks. Leveraging methods such as masked autoencoding and contrastive learning, these representations exhibit strong transfer to policy learning f…

Cited by 153Source ↗PDF ↗code ↗Scholar ↗
2023

Lexinvariant Language Models

Qian Huang, Eric Zelikman, Sarah Li Chen, Yuhuai Wu, Gregory Valiant, Percy Liang
NeurIPS 2023spotlight

Token embeddings, a mapping from discrete lexical symbols to continuous vectors, are at the heart of any language model (LM). However, lexical symbol meanings can also be determined and even redefined by their structural role in a long context. In this paper, we ask: is it possible for a language mo…

Cited by 3Source ↗PDF ↗Scholar ↗
2023

Out-of-Domain Robustness via Targeted Augmentations

Irena Gao, Shiori Sagawa, Pang Wei Koh, Tatsunori Hashimoto, Percy Liang
ICML 2023poster

Models trained on one set of domains often suffer performance drops on unseen domains, e.g., when wildlife monitoring models are deployed in new camera locations. In this work, we study principles for designing data augmentations for out-of-domain (OOD) generalization. In particular, we focus on rea…

Cited by 25Source ↗PDF ↗code ↗Scholar ↗
2023

PRODIGY: Enabling In-context Learning Over Graphs

Qian Huang, Hongyu Ren, Peng Chen, Gregor Kržmanc, Daniel Zeng, Percy Liang +1
NeurIPS 2023spotlight

In-context learning is the ability of a pretrained model to adapt to novel and diverse downstream tasks by conditioning on prompt examples, without optimizing any parameters. While large language models have demonstrated this ability, how in-context learning could be performed over graphs is unexpl…

Cited by 79Source ↗PDF ↗Scholar ↗
2023

Retrieval-Augmented Multimodal Language Modeling

Michihiro Yasunaga, Armen Aghajanyan, Weijia Shi, Richard James, Jure Leskovec, Percy Liang +3
ICML 2023poster

Recent multimodal models such as DALL-E and CM3 have achieved remarkable progress in text-to-image and image-to-text generation. However, these models store all their knowledge (e.g., the appearance of the Eiffel Tower) in the model parameters, requiring increasingly larger models and training data…

Cited by 159Source ↗PDF ↗Scholar ↗
2023

Surgical Fine-Tuning Improves Adaptation to Distribution Shifts

Yoonho Lee, Annie S Chen, Fahim Tajwar, Ananya Kumar, Huaxiu Yao, Percy Liang +1
ICLR 2023poster

A common approach to transfer learning under distribution shift is to fine-tune the last few layers of a pre-trained model, preserving learned features while also adapting to the new task. This paper shows that in such settings, selectively fine-tuning a subset of layers (which we term surgical fine…

Cited by 229Source ↗PDF ↗code ↗Scholar ↗
2023

Whose Opinions Do Language Models Reflect?

Shibani Santurkar, Esin Durmus, Faisal Ladhak, Cinoo Lee, Percy Liang, Tatsunori Hashimoto
ICML 2023oral

Language models (LMs) are increasingly being used in open-ended contexts, where the opinions they reflect in response to subjective queries can have a profound impact, both on user satisfaction, and shaping the views of society at large. We put forth a quantitative framework to investigate the opini…

Cited by 470Source ↗PDF ↗code ↗Scholar ↗
2022

An Explanation of In-context Learning as Implicit Bayesian Inference

Sang Michael Xie, Aditi Raghunathan, Percy Liang, Tengyu Ma
ICLR 2022poster

Large language models (LMs) such as GPT-3 have the surprising ability to do in-context learning, where the model learns to do a downstream task simply by conditioning on a prompt consisting of input-output examples. The LM learns from these examples without being explicitly pretrained to learn. Thus…

Cited by 795Source ↗PDF ↗code ↗Scholar ↗
2022

Calibrated ensembles can mitigate accuracy tradeoffs under distribution shift

Ananya Kumar, Tengyu Ma, Percy Liang, Aditi Raghunathan
UAI 2022poster

We often see undesirable tradeoffs in robust machine learning where out-of-distribution (OOD) accuracy is at odds with in-distribution (ID) accuracy. A robust classifier obtained via specialized techniques such as removing spurious features often has better OOD but worse ID accuracy compared to a st…

Cited by 36Source ↗PDF ↗Scholar ↗
2022

Connect, Not Collapse: Explaining Contrastive Learning for Unsupervised Domain Adaptation

Kendrick Shen, Robbie M Jones, Ananya Kumar, Sang Michael Xie, Jeff Z. Haochen, Tengyu Ma +1
ICML 2022oral

We consider unsupervised domain adaptation (UDA), where labeled data from a source domain (e.g., photos) and unlabeled data from a target domain (e.g., sketches) are used to learn a classifier for the target domain. Conventional UDA methods (e.g., domain adversarial training) learn domain-invariant…

Cited by 122Source ↗PDF ↗code ↗Scholar ↗
2022

Decentralized Training of Foundation Models in Heterogeneous Environments

Binhang Yuan, Yongjun He, Jared Quincy Davis, Tianyi Zhang, Tri Dao, Beidi Chen +3
NeurIPS 2022accept

Training foundation models, such as GPT-3 and PaLM, can be extremely expensive, often involving tens of thousands of GPUs running continuously for months. These models are typically trained in specialized clusters featuring fast, homogeneous interconnects and using carefully designed software system…

Cited by 105Source ↗PDF ↗code ↗Scholar ↗
2022

Deep Bidirectional Language-Knowledge Graph Pretraining

Michihiro Yasunaga, Antoine Bosselut, Hongyu Ren, Xikun Zhang, Christopher D Manning, Percy Liang +1
NeurIPS 2022accept

Pretraining a language model (LM) on text has been shown to help various downstream NLP tasks. Recent works show that a knowledge graph (KG) can complement text data, offering structured background knowledge that provides a useful scaffold for reasoning. However, these works are not pretrained to le…

Cited by 255Source ↗PDF ↗code ↗Scholar ↗
2022

Diffusion-LM Improves Controllable Text Generation

Xiang Lisa Li, John Thickstun, Ishaan Gulrajani, Percy Liang, Tatsunori Hashimoto
NeurIPS 2022accept

Controlling the behavior of language models (LMs) without re-training is a major open problem in natural language generation. While recent works have demonstrated successes on controlling simple sentence attributes (e.g., sentiment), there has been little progress on complex, fine-grained controls (…

Cited by 824Source ↗PDF ↗code ↗Scholar ↗
2022

Extending the WILDS Benchmark for Unsupervised Adaptation

Shiori Sagawa, Pang Wei Koh, Tony Lee, Irena Gao, Sang Michael Xie, Kendrick Shen +14
ICLR 2022oral

Machine learning systems deployed in the wild are often trained on a source distribution but deployed on a different target distribution. Unlabeled data can be a powerful point of leverage for mitigating these distribution shifts, as it is frequently much more available than labeled data and can oft…

Cited by 143Source ↗PDF ↗Scholar ↗
2022

Fine-Tuning can Distort Pretrained Features and Underperform Out-of-Distribution

Ananya Kumar, Aditi Raghunathan, Robbie Matthew Jones, Tengyu Ma, Percy Liang
ICLR 2022oral

When transferring a pretrained model to a downstream task, two popular methods are full fine-tuning (updating all the model parameters) and linear probing (updating only the last linear layer---the "head"). It is well known that fine-tuning leads to better accuracy in-distribution (ID). However, in…

Cited by 807Source ↗PDF ↗Scholar ↗
2022

GreaseLM: Graph REASoning Enhanced Language Models

Xikun Zhang, Antoine Bosselut, Michihiro Yasunaga, Hongyu Ren, Percy Liang, Christopher D Manning +1
ICLR 2022spotlight

Answering complex questions about textual narratives requires reasoning over both stated context and the world knowledge that underlies it. However, pretrained language models (LM), the foundation of most modern QA systems, do not robustly represent latent relationships between concepts, which is ne…

Cited by 0Source ↗PDF ↗Scholar ↗
2022

Improving Self-Supervised Learning by Characterizing Idealized Representations

Yann Dubois, Stefano Ermon, Tatsunori Hashimoto, Percy Liang
NeurIPS 2022accept

Despite the empirical successes of self-supervised learning (SSL) methods, it is unclear what characteristics of their representations lead to high downstream accuracies. In this work, we characterize properties that SSL representations should ideally satisfy. Specifically, we prove necessary and su…

Cited by 42Source ↗PDF ↗code ↗Scholar ↗
2022

Large Language Models Can Be Strong Differentially Private Learners

Xuechen Li, Florian Tramer, Percy Liang, Tatsunori Hashimoto
ICLR 2022oral

Differentially Private (DP) learning has seen limited success for building large deep learning models of text, and straightforward attempts at applying Differentially Private Stochastic Gradient Descent (DP-SGD) to NLP tasks have resulted in large performance drops and high computational overhead. W…

Cited by 415Source ↗PDF ↗code ↗Scholar ↗
2022

Picking on the Same Person: Does Algorithmic Monoculture lead to Outcome Homogenization?

Rishi Bommasani, Kathleen Creel, Ananya Kumar, Dan Jurafsky, Percy Liang
NeurIPS 2022accept

As the scope of machine learning broadens, we observe a recurring theme of *algorithmic monoculture*: the same systems, or systems that share components (e.g. datasets, models), are deployed by multiple decision-makers. While sharing offers advantages like amortizing effort, it also has risks. We…

Cited by 100Source ↗PDF ↗Scholar ↗
2022

What Can Transformers Learn In-Context? A Case Study of Simple Function Classes

Shivam Garg, Dimitris Tsipras, Percy Liang, Gregory Valiant
NeurIPS 2022accept

In-context learning is the ability of a model to condition on a prompt sequence consisting of in-context examples (input-output pairs corresponding to some task) along with a new query input, and generate the corresponding output. Crucially, in-context learning happens only at inference time without…

Cited by 491Source ↗PDF ↗code ↗Scholar ↗
2021

Accuracy on the Line: on the Strong Correlation Between Out-of-Distribution and In-Distribution Generalization

John P Miller, Rohan Taori, Aditi Raghunathan, Shiori Sagawa, Pang Wei Koh, Vaishaal Shankar +3
ICML 2021spotlight

For machine learning systems to be reliable, we must understand their performance in unseen, out- of-distribution environments. In this paper, we empirically show that out-of-distribution performance is strongly correlated with in-distribution performance for a wide range of models and distribution…

Cited by 333Source ↗PDF ↗Scholar ↗
2021

Catformer: Designing Stable Transformers via Sensitivity Analysis

Jared Q Davis, Albert Gu, Krzysztof Choromanski, Tri Dao, Christopher Re, Chelsea Finn +1
ICML 2021spotlight

Transformer architectures are widely used, but training them is non-trivial, requiring custom learning rate schedules, scaling terms, residual connections, careful placement of submodules such as normalization, and so on. In this paper, we improve upon recent analysis of Transformers and formalize a…

Cited by 22Source ↗PDF ↗Scholar ↗
2021

Composed Fine-Tuning: Freezing Pre-Trained Denoising Autoencoders for Improved Generalization

Sang Michael Xie, Tengyu Ma, Percy Liang
ICML 2021oral

We focus on prediction problems with structured outputs that are subject to output validity constraints, e.g. pseudocode-to-code translation where the code must compile. While labeled input-output pairs are expensive to obtain, "unlabeled" outputs, i.e. outputs without corresponding inputs, are free…

Cited by 18Source ↗PDF ↗Scholar ↗
2021

Conditional probing: measuring usable information beyond a baseline

John Hewitt, Kawin Ethayarajh, Percy Liang, Christopher Manning
EMNLP 2021main

Probing experiments investigate the extent to which neural representations make properties—like part-of-speech—predictable. One suggests that a representation encodes a property if probing that representation produces higher accuracy than probing a baseline representation like non-contextual word em…

Cited by 66Source ↗PDF ↗code ↗Scholar ↗
2021

Decoupling Exploration and Exploitation for Meta-Reinforcement Learning without Sacrifices

Evan Z Liu, Aditi Raghunathan, Percy Liang, Chelsea Finn
ICML 2021spotlight

The goal of meta-reinforcement learning (meta-RL) is to build agents that can quickly learn new tasks by leveraging prior experience on related tasks. Learning a new task often requires both exploring to gather task-relevant information and exploiting this information to solve the task. In principle…

Cited by 90Source ↗PDF ↗code ↗Scholar ↗
2021

In-N-Out: Pre-Training and Self-Training using Auxiliary Information for Out-of-Distribution Robustness

Sang Michael Xie, Ananya Kumar, Robbie Jones, Fereshte Khani, Tengyu Ma, Percy Liang
ICLR 2021poster

Consider a prediction setting with few in-distribution labeled examples and many unlabeled examples both in- and out-of-distribution (OOD). The goal is to learn a model which performs well both in-distribution and OOD. In these settings, auxiliary information is often cheaply available for every inp…

Cited by 71Source ↗PDF ↗Scholar ↗
2021

Just Train Twice: Improving Group Robustness without Training Group Information

Evan Z Liu, Behzad Haghgoo, Annie S Chen, Aditi Raghunathan, Pang Wei Koh, Shiori Sagawa +2
ICML 2021oral

Standard training via empirical risk minimization (ERM) can produce models that achieve low error on average but high error on minority groups, especially in the presence of spurious correlations between the input and label. Prior approaches to this problem, like group distributionally robust optimi…

Cited by 630Source ↗PDF ↗code ↗Scholar ↗
2021

LM-Critic: Language Models for Unsupervised Grammatical Error Correction

Michihiro Yasunaga, Jure Leskovec, Percy Liang
EMNLP 2021main

Grammatical error correction (GEC) requires a set of labeled ungrammatical / grammatical sentence pairs for training, but obtaining such annotation can be prohibitively expensive. Recently, the Break-It-Fix-It (BIFI) framework has demonstrated strong results on learning to repair a broken program wi…

Cited by 55Source ↗PDF ↗code ↗Scholar ↗
2021

QA-GNN: Reasoning with Language Models and Knowledge Graphs for Question Answering

Michihiro Yasunaga, Hongyu Ren, Antoine Bosselut, Percy Liang, Jure Leskovec
NAACL 2021long

The problem of answering questions using knowledge from pre-trained language models (LMs) and knowledge graphs (KGs) presents two challenges: given a QA context (question and answer choice), methods need to (i) identify relevant knowledge from large KGs, and (ii) perform joint reasoning over the QA…

Cited by 689Source ↗PDF ↗code ↗Scholar ↗
2021

Selective Classification Can Magnify Disparities Across Groups

Erik Jones, Shiori Sagawa, Pang Wei Koh, Ananya Kumar, Percy Liang
ICLR 2021poster

Selective classification, in which models can abstain on uncertain predictions, is a natural approach to improving accuracy in settings where errors are costly but abstentions are manageable. In this paper, we find that while selective classification can improve average accuracies, it can simultaneo…

Cited by 62Source ↗PDF ↗Scholar ↗
2021

Swords: A Benchmark for Lexical Substitution with Improved Data Coverage and Quality

Mina Lee, Chris Donahue, Robin Jia, Alexander Iyabor, Percy Liang
NAACL 2021long

We release a new benchmark for lexical substitution, the task of finding appropriate substitutes for a target word in a context. For writing, lexical substitution systems can assist humans by suggesting words that humans cannot easily think of. However, existing benchmarks depend on human recall as…

Cited by 22Source ↗PDF ↗code ↗Scholar ↗
2021

WILDS: A Benchmark of in-the-Wild Distribution Shifts

Pang Wei Koh, Shiori Sagawa, Henrik Marklund, Sang Michael Xie, Marvin Zhang, Akshay Balsubramani +17
ICML 2021oral

Distribution shifts—where the training distribution differs from the test distribution—can substantially degrade the accuracy of machine learning (ML) systems deployed in the wild. Despite their ubiquity in the real-world deployments, these distribution shifts are under-represented in the datasets w…

Cited by 1715Source ↗PDF ↗code ↗Scholar ↗
2020

An Investigation of Why Overparameterization Exacerbates Spurious Correlations

Shiori Sagawa, Aditi Raghunathan, Pang Wei Koh, Percy Liang
ICML 2020poster

We study why overparameterization—increasing model size well beyond the point of zero training error—can hurt test error on minority groups despite improving average test error when there are spurious correlations in the data. Through simulations and experiments on two image datasets, we identify tw…

Cited by 444Source ↗PDF ↗code ↗Scholar ↗
2020

Concept Bottleneck Models

Pang Wei Koh, Thao Nguyen, Yew Siang Tang, Stephen Mussmann, Emma Pierson, Been Kim +1
ICML 2020poster

We seek to learn models that we can interact with using high-level concepts: if the model did not think there was a bone spur in the x-ray, would it still predict severe arthritis? State-of-the-art models today do not typically support the manipulation of concepts like "the existence of bone spurs",…

Cited by 1035Source ↗PDF ↗code ↗Scholar ↗
2020

Enabling certification of verification-agnostic networks via memory-efficient semidefinite programming

Sumanth Dathathri, Krishnamurthy Dvijotham, Alexey Kurakin, Aditi Raghunathan, Jonathan Uesato, Rudy R Bunel +5
NeurIPS 2020poster

Convex relaxations have emerged as a promising approach for verifying properties of neural networks, but widely used using Linear Programming (LP) relaxations only provide meaningful certificates when networks are specifically trained to facilitate verification. This precludes many important applica…

Cited by 130Source ↗PDF ↗code ↗Scholar ↗
2020

Selection via Proxy: Efficient Data Selection for Deep Learning

Cody Coleman, Christopher Yeh, Stephen Mussmann, Baharan Mirzasoleiman, Peter Bailis, Percy Liang +2
ICLR 2020poster

Data selection methods, such as active learning and core-set selection, are useful tools for machine learning on large datasets. However, they can be prohibitively expensive to apply in deep learning because they depend on feature representations that need to be learned. In this work, we show that w…

Cited by 391Source ↗code ↗Scholar ↗
2020

Strategies for Pre-training Graph Neural Networks

Weihua Hu*, Bowen Liu*, Joseph Gomes, Marinka Zitnik, Percy Liang, Vijay Pande +1
ICLR 2020spotlight

Many applications of machine learning require a model to make accurate pre-dictions on test examples that are distributionally different from training ones, while task-specific labels are scarce during training. An effective approach to this challenge is to pre-train a model on related tasks where d…

Cited by 1823Source ↗code ↗Scholar ↗
2020

Understanding and Mitigating the Tradeoff between Robustness and Accuracy

Aditi Raghunathan, Sang Michael Xie, Fanny Yang, John Duchi, Percy Liang
ICML 2020poster

Adversarial training augments the training set with perturbations to improve the robust error (over worst-case perturbations), but it often leads to an increase in the standard error (on unperturbed test inputs). Previous explanations for this tradeoff rely on the assumption that no predictor in the…

Cited by 286Source ↗PDF ↗Scholar ↗
2019

Inferring Multidimensional Rates of Aging from Cross-Sectional Data

Emma Pierson, Pang Wei Koh, Tatsunori Hashimoto, Daphne Koller, Jure Leskovec, Nick Eriksson +1
AISTATS 2019poster

Modeling how individuals evolve over time is a fundamental problem in the natural and social sciences. However, existing datasets are often cross-sectional with each individual observed only once, making it impossible to apply traditional time-series methods. Motivated by the study of human aging, w…

Cited by 32Source ↗PDF ↗code ↗Scholar ↗
2019

Learning a SAT Solver from Single-Bit Supervision

Daniel Selsam, Matthew Lamm, Benedikt B\"{u}nz, Percy Liang, Leonardo de Moura, David L. Dill
ICLR 2019poster

We present NeuroSAT, a message passing neural network that learns to solve SAT problems after only being trained as a classifier to predict satisfiability. Although it is not competitive with state-of-the-art SAT solvers, NeuroSAT can solve problems that are substantially larger and more difficult…

Cited by 541Source ↗PDF ↗Scholar ↗
2019

On the Accuracy of Influence Functions for Measuring Group Effects

Pang Wei W Koh, Kai-Siang Ang, Hubert Teo, Percy Liang
NeurIPS 2019poster

Influence functions estimate the effect of removing a training point on a model without the need to retrain. They are based on a first-order Taylor approximation that is guaranteed to be accurate for sufficiently small changes to the model, and so are commonly used to study the effect of individual…

Cited by 243Source ↗PDF ↗code ↗Scholar ↗
2019

SPoC: Search-based Pseudocode to Code

Sumith Kulal, Panupong Pasupat, Kartik Chandra, Mina Lee, Oded Padon, Alex Aiken +1
NeurIPS 2019poster

We consider the task of mapping pseudocode to executable code, assuming a one-to-one correspondence between lines of pseudocode and lines of code. Given test cases as a mechanism to validate programs, we search over the space of possible translations of the pseudocode to find a program that compiles…

Cited by 256Source ↗PDF ↗Scholar ↗
2019

Unlabeled Data Improves Adversarial Robustness

Yair Carmon, Aditi Raghunathan, Ludwig Schmidt, John C. Duchi, Percy Liang
NeurIPS 2019poster

We demonstrate, theoretically and empirically, that adversarial robustness can significantly benefit from semisupervised learning. Theoretically, we revisit the simple Gaussian model of Schmidt et al. that shows a sample complexity gap between standard and robust classification. We prove that unlab…

Cited by 897Source ↗PDF ↗code ↗Scholar ↗
2018

A Retrieve-and-Edit Framework for Predicting Structured Outputs

Tatsunori B Hashimoto, Kelvin Guu, Yonatan Oren, Percy Liang
NeurIPS 2018oral

For the task of generating complex outputs such as source code, editing existing outputs can be easier than generating complex outputs from scratch. With this motivation, we propose an approach that first retrieves a training example based on the input (e.g., natural language description) and then e…

Cited by 193Source ↗PDF ↗Scholar ↗
2018

Fairness Without Demographics in Repeated Loss Minimization

Tatsunori Hashimoto, Megha Srivastava, Hongseok Namkoong, Percy Liang
ICML 2018oral

Machine learning models (e.g., speech recognizers) trained on average loss suffer from representation disparity—minority groups (e.g., non-native speakers) carry less weight in the training objective, and thus tend to suffer higher loss. Worse, as model accuracy affects user retention, a minority gr…

Cited by 757Source ↗PDF ↗Scholar ↗
2018

Reinforcement Learning on Web Interfaces using Workflow-Guided Exploration

Evan Zheran Liu, Kelvin Guu, Panupong Pasupat, Tianlin Shi, Percy Liang
ICLR 2018poster

Reinforcement learning (RL) agents improve through trial-and-error, but when reward is sparse and the agent cannot discover successful action sequences, learning stagnates. This has been a notable problem in training deep RL agents to perform web-based tasks, such as booking flights or replying to e…

Cited by 234Source ↗PDF ↗code ↗Scholar ↗
2018

Semidefinite relaxations for certifying robustness to adversarial examples

Aditi Raghunathan, Jacob Steinhardt, Percy Liang
NeurIPS 2018poster

Despite their impressive performance on diverse tasks, neural networks fail catastrophically in the presence of adversarial inputs—imperceptibly but adversarially perturbed versions of natural inputs. We have witnessed an arms race between defenders who attempt to train robust networks and attackers…

Cited by 538Source ↗PDF ↗Scholar ↗
2018

Uncertainty Sampling is Preconditioned Stochastic Gradient Descent on Zero-One Loss

Stephen Mussmann, Percy Liang
NeurIPS 2018poster

Uncertainty sampling, a popular active learning algorithm, is used to reduce the amount of data required to learn a classifier, but it has been observed in practice to converge to different parameters depending on the initialization and sometimes to even better parameters than standard training on a…

Cited by 23Source ↗PDF ↗Scholar ↗
2017

Developing Bug-Free Machine Learning Systems With Formal Mathematics

Daniel Selsam, Percy Liang, David L. Dill
ICML 2017poster

Noisy data, non-convex objectives, model misspecification, and numerical instability can all cause undesired behaviors in machine learning systems. As a result, detecting actual implementation errors can be extremely difficult. We demonstrate a methodology in which developers use an interactive proo…

Cited by 67Source ↗PDF ↗code ↗Scholar ↗
2017

Unsupervised Transformation Learning via Convex Relaxations

Tatsunori B Hashimoto, Percy Liang, John C. Duchi
NeurIPS 2017poster

Our goal is to extract meaningful transformations from raw images, such as varying the thickness of lines in handwriting or the lighting in a portrait. We propose an unsupervised approach to learn such transformations by attempting to reconstruct an image from a linear combination of transformations…

Cited by 12Source ↗PDF ↗Scholar ↗
2017

World of Bits: An Open-Domain Platform for Web-Based Agents

Tianlin Shi, Andrej Karpathy, Linxi Fan, Jonathan Hernandez, Percy Liang
ICML 2017poster

While simulated game environments have greatly accelerated research in reinforcement learning, existing environments lack the open-domain realism of tasks in computer vision or natural language processing, which operate on artifacts created by humans in natural, organic settings. To foster reinforce…

Cited by 244Source ↗PDF ↗Scholar ↗
2016

Estimation from Indirect Supervision with Linear Moments

Aditi Raghunathan, Roy Frostig, John Duchi, Percy Liang
ICML 2016poster

In structured prediction problems where we have indirect supervision of the output, maximum marginal likelihood faces two computational obstacles: non-convexity of the objective and intractability of even a single gradient computation. In this paper, we bypass both obstacles for a class of what we c…

Cited by 16Source ↗PDF ↗Scholar ↗
2015

On-the-Job Learning with Bayesian Decision Theory

Keenon Werling, Arun Tejasvi Chaganty, Percy Liang, Christopher D. Manning
NeurIPS 2015spotlight

Our goal is to deploy a high-accuracy system starting with zero training examples. We consider an “on-the-job” setting, where as inputs arrive, we use real-time crowdsourcing to resolve uncertainty where needed and output our prediction when confident. As the model improves over time, the reliance o…

Cited by 31Source ↗PDF ↗code ↗Scholar ↗
2015

Reified Context Models

Jacob Steinhardt, Percy Liang
ICML 2015poster

A classic tension exists between exact inference in a simple model and approximate inference in a complex model. The latter offers expressivity and thus accuracy, but the former provides coverage of the space, an important property for confidence estimation and learning with indirect supervision. In…

Cited by 5Source ↗PDF ↗Scholar ↗