Ananya Kumar

Aaron Jaech, Adam Kalai, Adam Lerer et al. · openai

The o1 model series is trained with large-scale reinforcement learning to reason using chain of thought. These advanced reasoning capabilities provide new avenues for improving the safety and robustness of our models. In particular, our models can reason about our safety policies in context when responding to potentially unsafe prompts, through deliberative alignment. This leads to state-of-the-art performance on certain benchmarks for risks such as generating illicit advice, choosing stereotyped responses, and succumbing to known jailbreaks. Training models to incorporate a chain of thought before answering has the potential to unlock substantial benefits, while also increasing potential risks that stem from heightened intelligence. Our results underscore the need for building robust alignment methods, extensively stress-testing their efficacy, and maintaining meticulous risk management protocols. This report outlines the safety work carried out for the OpenAI o1 and OpenAI o1-mini models, including safety evaluations, external red teaming, and Preparedness Framework evaluations.

Aaron Hurst, Adam Lerer, Adam P. Goucher et al. · openai

GPT-4o is an autoregressive omni model that accepts as input any combination of text, audio, image, and video, and generates any combination of text, audio, and image outputs. It's trained end-to-end across text, vision, and audio, meaning all inputs and outputs are processed by the same neural network. GPT-4o can respond to audio inputs in as little as 232 milliseconds, with an average of 320 milliseconds, which is similar to human response time in conversation. It matches GPT-4 Turbo performance on text in English and code, with significant improvement on text in non-English languages, while also being much faster and 50\% cheaper in the API. GPT-4o is especially better at vision and audio understanding compared to existing models. In line with our commitment to building AI safely and consistent with our voluntary commitments to the White House, we are sharing the GPT-4o System Card, which includes our Preparedness Framework evaluations. In this System Card, we provide a detailed look at GPT-4o's capabilities, limitations, and safety evaluations across multiple categories, focusing on speech-to-speech while also evaluating text and image capabilities, and measures we've implemented to ensure the model is safe and aligned. We also include third-party assessments on dangerous capabilities, as well as discussion of potential societal impacts of GPT-4o's text and vision capabilities.

Sachin Goyal, Ananya Kumar, Sankalp Garg et al. · apple-ml, cmu

Finetuning image-text models such as CLIP achieves state-of-the-art accuracies on a variety of benchmarks. However, recent works like WiseFT (Wortsman et al., 2021) and LP-FT (Kumar et al., 2022) have shown that even subtle differences in the finetuning process can lead to surprisingly large differences in the final performance, both for in-distribution (ID) and out-of-distribution (OOD) data. In this work, we show that a natural and simple approach of mimicking contrastive pretraining consistently outperforms alternative finetuning approaches. Specifically, we cast downstream class labels as text prompts and continue optimizing the contrastive loss between image embeddings and class-descriptive prompt embeddings (contrastive finetuning). Our method consistently outperforms baselines across 7 distribution shifts, 6 transfer learning, and 3 few-shot learning benchmarks. On WILDS-iWILDCam, our proposed approach FLYP outperforms the top of the leaderboard by $2.3\%$ ID and $2.7\%$ OOD, giving the highest reported accuracy. Averaged across 7 OOD datasets (2 WILDS and 5 ImageNet associated shifts), FLYP gives gains of $4.2\%$ OOD over standard finetuning and outperforms the current state of the art (LP-FT) by more than $1\%$ both ID and OOD. Similarly, on 3 few-shot learning benchmarks, our approach gives gains up to $4.6\%$ over standard finetuning and $4.4\%$ over the state of the art. In total, these benchmarks establish contrastive finetuning as a simple, intuitive, and state-of-the-art approach for supervised finetuning of image-text models like CLIP. Code is available at https://github.com/locuslab/FLYP.

Rishi Bommasani, Kathleen A. Creel, Ananya Kumar et al. · stanford

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. training data), are deployed by multiple decision-makers. While sharing offers clear advantages (e.g. amortizing costs), does it bear risks? We introduce and formalize one such risk, outcome homogenization: the extent to which particular individuals or groups experience negative outcomes from all decision-makers. If the same individuals or groups exclusively experience undesirable outcomes, this may institutionalize systemic exclusion and reinscribe social hierarchy. To relate algorithmic monoculture and outcome homogenization, we propose the component-sharing hypothesis: if decision-makers share components like training data or specific models, then they will produce more homogeneous outcomes. We test this hypothesis on algorithmic fairness benchmarks, demonstrating that sharing training data reliably exacerbates homogenization, with individual-level effects generally exceeding group-level effects. Further, given the dominant paradigm in AI of foundation models, i.e. models that can be adapted for myriad downstream tasks, we test whether model sharing homogenizes outcomes across tasks. We observe mixed results: we find that for both vision and language settings, the specific methods for adapting a foundation model significantly influence the degree of outcome homogenization. We conclude with philosophical analyses of and societal challenges for outcome homogenization, with an eye towards implications for deployed machine learning systems.

Kendrick Shen, Robbie Jones, Ananya Kumar et al. · stanford

We consider unsupervised domain adaptation (UDA), where labeled data from a source domain (e.g., photographs) 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 features to improve generalization to the target domain. In this paper, we show that contrastive pre-training, which learns features on unlabeled source and target data and then fine-tunes on labeled source data, is competitive with strong UDA methods. However, we find that contrastive pre-training does not learn domain-invariant features, diverging from conventional UDA intuitions. We show theoretically that contrastive pre-training can learn features that vary subtantially across domains but still generalize to the target domain, by disentangling domain and class information. Our results suggest that domain invariance is not necessary for UDA. We empirically validate our theory on benchmark vision datasets.

Aaditya Singh, Adam Fry, Adam Perelman et al. · berkeley, mila

This is the system card published alongside the OpenAI GPT-5 launch, August 2025. GPT-5 is a unified system with a smart and fast model that answers most questions, a deeper reasoning model for harder problems, and a real-time router that quickly decides which model to use based on conversation type, complexity, tool needs, and explicit intent (for example, if you say 'think hard about this' in the prompt). The router is continuously trained on real signals, including when users switch models, preference rates for responses, and measured correctness, improving over time. Once usage limits are reached, a mini version of each model handles remaining queries. This system card focuses primarily on gpt-5-thinking and gpt-5-main, while evaluations for other models are available in the appendix. The GPT-5 system not only outperforms previous models on benchmarks and answers questions more quickly, but -- more importantly -- is more useful for real-world queries. We've made significant advances in reducing hallucinations, improving instruction following, and minimizing sycophancy, and have leveled up GPT-5's performance in three of ChatGPT's most common uses: writing, coding, and health. All of the GPT-5 models additionally feature safe-completions, our latest approach to safety training to prevent disallowed content. Similarly to ChatGPT agent, we have decided to treat gpt-5-thinking as High capability in the Biological and Chemical domain under our Preparedness Framework, activating the associated safeguards. While we do not have definitive evidence that this model could meaningfully help a novice to create severe biological harm -- our defined threshold for High capability -- we have chosen to take a precautionary approach.

Nelson F. Liu, Ananya Kumar, Percy Liang et al. · stanford

Recent results in image classification and extractive question answering have observed that pre-trained models trained on less in-distribution data have better out-of-distribution performance. However, it is unclear how broadly these trends hold. We conduct a large empirical study across three tasks, three broadly-applicable modeling interventions (increasing model size, using a different adaptation method, and pre-training on more data), and 14 diverse datasets to investigate the relationship between sample efficiency (amount of data needed to reach a given ID accuracy) and robustness (how models fare on OOD evaluation). We find that higher sample efficiency is only correlated with better average OOD robustness on some modeling interventions and tasks, but not others. On individual datasets, models with lower sample efficiency can even be more robust. These results suggest that general-purpose methods for improving sample efficiency are unlikely to yield universal OOD robustness improvements, since such improvements are highly dataset- and task-dependent. Even in an era of large, multi-purpose pretrained models, task-specific decisions may often be necessary for OOD generalization.

Yoonho Lee, Annie S. Chen, Fahim Tajwar et al.

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-tuning) matches or outperforms commonly used fine-tuning approaches. Moreover, the type of distribution shift influences which subset is more effective to tune: for example, for image corruptions, fine-tuning only the first few layers works best. We validate our findings systematically across seven real-world data tasks spanning three types of distribution shifts. Theoretically, we prove that for two-layer neural networks in an idealized setting, first-layer tuning can outperform fine-tuning all layers. Intuitively, fine-tuning more parameters on a small target dataset can cause information learned during pre-training to be forgotten, and the relevant information depends on the type of shift.

Jeff Z. HaoChen, Colin Wei, Ananya Kumar et al. · stanford

Contrastive learning is a highly effective method for learning representations from unlabeled data. Recent works show that contrastive representations can transfer across domains, leading to simple state-of-the-art algorithms for unsupervised domain adaptation. In particular, a linear classifier trained to separate the representations on the source domain can also predict classes on the target domain accurately, even though the representations of the two domains are far from each other. We refer to this phenomenon as linear transferability. This paper analyzes when and why contrastive representations exhibit linear transferability in a general unsupervised domain adaptation setting. We prove that linear transferability can occur when data from the same class in different domains (e.g., photo dogs and cartoon dogs) are more related with each other than data from different classes in different domains (e.g., photo dogs and cartoon cats) are. Our analyses are in a realistic regime where the source and target domains can have unbounded density ratios and be weakly related, and they have distant representations across domains.

Ananya Kumar, Ruoqi Shen, Sebastien Bubeck et al. · microsoft-research

SGD and AdamW are the two most used optimizers for fine-tuning large neural networks in computer vision. When the two methods perform the same, SGD is preferable because it uses less memory (12 bytes/parameter with momentum and 8 bytes/parameter without) than AdamW (16 bytes/parameter). However, on a suite of downstream tasks, especially those with distribution shifts, we find that fine-tuning with AdamW performs substantially better than SGD on modern Vision Transformer and ConvNeXt models. We find that large gaps in performance between SGD and AdamW occur when the fine-tuning gradients in the first "embedding" layer are much larger than in the rest of the model. Our analysis suggests an easy fix that works consistently across datasets and models: freezing the embedding layer (less than 1% of the parameters) leads to SGD with or without momentum performing slightly better than AdamW while using less memory (e.g., on ViT-L, SGD uses 33% less GPU memory). Our insights result in state-of-the-art accuracies on five popular distribution shift benchmarks: WILDS-FMoW, WILDS-Camelyon, BREEDS-Living-17, Waterbirds, and DomainNet.

Percy Liang, Rishi Bommasani, Tony Lee et al. · stanford

Language models (LMs) are becoming the foundation for almost all major language technologies, but their capabilities, limitations, and risks are not well understood. We present Holistic Evaluation of Language Models (HELM) to improve the transparency of language models. First, we taxonomize the vast space of potential scenarios (i.e. use cases) and metrics (i.e. desiderata) that are of interest for LMs. Then we select a broad subset based on coverage and feasibility, noting what's missing or underrepresented (e.g. question answering for neglected English dialects, metrics for trustworthiness). Second, we adopt a multi-metric approach: We measure 7 metrics (accuracy, calibration, robustness, fairness, bias, toxicity, and efficiency) for each of 16 core scenarios when possible (87.5% of the time). This ensures metrics beyond accuracy don't fall to the wayside, and that trade-offs are clearly exposed. We also perform 7 targeted evaluations, based on 26 targeted scenarios, to analyze specific aspects (e.g. reasoning, disinformation). Third, we conduct a large-scale evaluation of 30 prominent language models (spanning open, limited-access, and closed models) on all 42 scenarios, 21 of which were not previously used in mainstream LM evaluation. Prior to HELM, models on average were evaluated on just 17.9% of the core HELM scenarios, with some prominent models not sharing a single scenario in common. We improve this to 96.0%: now all 30 models have been densely benchmarked on the same core scenarios and metrics under standardized conditions. Our evaluation surfaces 25 top-level findings. For full transparency, we release all raw model prompts and completions publicly for further analysis, as well as a general modular toolkit. We intend for HELM to be a living benchmark for the community, continuously updated with new scenarios, metrics, and models.

Ananya Kumar, Tengyu Ma, Percy Liang et al.

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 standard classifier trained via ERM. In this paper, we find that ID-calibrated ensembles -- where we simply ensemble the standard and robust models after calibrating on only ID data -- outperforms prior state-of-the-art (based on self-training) on both ID and OOD accuracy. On eleven natural distribution shift datasets, ID-calibrated ensembles obtain the best of both worlds: strong ID accuracy and OOD accuracy. We analyze this method in stylized settings, and identify two important conditions for ensembles to perform well both ID and OOD: (1) we need to calibrate the standard and robust models (on ID data, because OOD data is unavailable), (2) OOD has no anticorrelated spurious features.

Caroline Choi, Fahim Tajwar, Yoonho Lee et al.

In safety-critical applications of machine learning, it is often desirable for a model to be conservative, abstaining from making predictions on unknown inputs which are not well-represented in the training data. However, detecting unknown examples is challenging, as it is impossible to anticipate all potential inputs at test time. To address this, prior work (Hendrycks et al., 2018) minimizes model confidence on an auxiliary outlier dataset carefully curated to be disjoint from the training distribution. We theoretically analyze the choice of auxiliary dataset for confidence minimization, revealing two actionable insights: (1) if the auxiliary set contains unknown examples similar to those seen at test time, confidence minimization leads to provable detection of unknown test examples, and (2) if the first condition is satisfied, it is unnecessary to filter out known examples for out-of-distribution (OOD) detection. Motivated by these guidelines, we propose the Data-Driven Confidence Minimization (DCM) framework, which minimizes confidence on an uncertainty dataset. We apply DCM to two problem settings in which conservative prediction is paramount -- selective classification and OOD detection -- and provide a realistic way to gather uncertainty data for each setting. In our experiments, DCM consistently outperforms existing selective classification approaches on 4 datasets when tested on unseen distributions and outperforms state-of-the-art OOD detection methods on 12 ID-OOD dataset pairs, reducing FPR (at TPR $95\%$) by $6.3\%$ and $58.1\%$ on CIFAR-10 and CIFAR-100 compared to Outlier Exposure.

Michael Sun, Ananya Kumar, Divyam Madaan et al.

We consider the continual representation learning setting: sequentially pretrain a model $M'$ on tasks $T_1, \ldots, T_T$, and then adapt $M'$ on a small amount of data from each task $T_i$ to check if it has forgotten information from old tasks. Under a kNN adaptation protocol, prior work shows that continual learning methods improve forgetting over naive training (SGD). In reality, practitioners do not use kNN classifiers -- they use the adaptation method that works best (e.g., fine-tuning) -- here, we find that strong continual learning baselines do worse than naive training. Interestingly, we find that a method from the transfer learning community (LP-FT) outperforms naive training and the other continual learning methods. Even with standard kNN evaluation protocols, LP-FT performs comparably with strong continual learning methods (while being simpler and requiring less memory) on three standard benchmarks: sequential CIFAR-10, CIFAR-100, and TinyImageNet. LP-FT also reduces forgetting in a real world satellite remote sensing dataset (FMoW), and a variant of LP-FT gets state-of-the-art accuracies on an NLP continual learning benchmark.

Vaishnavi Shrivastava, Percy Liang, Ananya Kumar

To maintain user trust, large language models (LLMs) should signal low confidence on examples where they are incorrect, instead of misleading the user. The standard approach of estimating confidence is to use the softmax probabilities of these models, but as of November 2023, state-of-the-art LLMs such as GPT-4 and Claude-v1.3 do not provide access to these probabilities. We first study eliciting confidence linguistically -- asking an LLM for its confidence in its answer -- which performs reasonably (80.5% AUC on GPT-4 averaged across 12 question-answering datasets -- 7% above a random baseline) but leaves room for improvement. We then explore using a surrogate confidence model -- using a model where we do have probabilities to evaluate the original model's confidence in a given question. Surprisingly, even though these probabilities come from a different and often weaker model, this method leads to higher AUC than linguistic confidences on 9 out of 12 datasets. Our best method composing linguistic confidences and surrogate model probabilities gives state-of-the-art confidence estimates on all 12 datasets (84.6% average AUC on GPT-4).

Yun-Shiuan Chuang, Yi Wu, Dhruv Gupta et al.

Adapting pre-trained language models (PLMs) for time-series text classification amidst evolving domain shifts (EDS) is critical for maintaining accuracy in applications like stance detection. This study benchmarks the effectiveness of evolving domain adaptation (EDA) strategies, notably self-training, domain-adversarial training, and domain-adaptive pretraining, with a focus on an incremental self-training method. Our analysis across various datasets reveals that this incremental method excels at adapting PLMs to EDS, outperforming traditional domain adaptation techniques. These findings highlight the importance of continually updating PLMs to ensure their effectiveness in real-world applications, paving the way for future research into PLM robustness against the natural temporal evolution of language.

Alexander C. Li, Ananya Kumar, Deepak Pathak

Discriminative approaches to classification often learn shortcuts that hold in-distribution but fail even under minor distribution shift. This failure mode stems from an overreliance on features that are spuriously correlated with the label. We show that generative classifiers, which use class-conditional generative models, can avoid this issue by modeling all features, both core and spurious, instead of mainly spurious ones. These generative classifiers are simple to train, avoiding the need for specialized augmentations, strong regularization, extra hyperparameters, or knowledge of the specific spurious correlations to avoid. We find that diffusion-based and autoregressive generative classifiers achieve state-of-the-art performance on five standard image and text distribution shift benchmarks and reduce the impact of spurious correlations in realistic applications, such as medical or satellite datasets. Finally, we carefully analyze a Gaussian toy setting to understand the inductive biases of generative classifiers, as well as the data properties that determine when generative classifiers outperform discriminative ones.

Shiori Sagawa, Pang Wei Koh, Tony Lee et al.

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 often be obtained from distributions beyond the source distribution as well. However, existing distribution shift benchmarks with unlabeled data do not reflect the breadth of scenarios that arise in real-world applications. In this work, we present the WILDS 2.0 update, which extends 8 of the 10 datasets in the WILDS benchmark of distribution shifts to include curated unlabeled data that would be realistically obtainable in deployment. These datasets span a wide range of applications (from histology to wildlife conservation), tasks (classification, regression, and detection), and modalities (photos, satellite images, microscope slides, text, molecular graphs). The update maintains consistency with the original WILDS benchmark by using identical labeled training, validation, and test sets, as well as the evaluation metrics. On these datasets, we systematically benchmark state-of-the-art methods that leverage unlabeled data, including domain-invariant, self-training, and self-supervised methods, and show that their success on WILDS is limited. To facilitate method development and evaluation, we provide an open-source package that automates data loading and contains all of the model architectures and methods used in this paper. Code and leaderboards are available at https://wilds.stanford.edu.

Vaishnavi Shrivastava, Ananya Kumar, Percy Liang

Language models (LMs) should provide reliable confidence estimates to help users detect mistakes in their outputs and defer to human experts when necessary. Asking a language model to assess its confidence ("Score your confidence from 0-1.") is a natural way of evaluating its uncertainty. However, models struggle to provide absolute assessments of confidence (i.e. judging confidence in answering a question independent of other questions) and the coarse-grained scores they produce are not useful for evaluating the correctness of their answers. We propose relative confidence estimation, where we match up questions against each other and ask the model to make relative judgments of confidence ("Which question are you more confident in answering correctly?"). Treating each question as a "player" in a series of matchups against other questions and the model's preferences as match outcomes, we can use rank aggregation methods like Elo rating and Bradley-Terry to translate the model's confidence preferences into confidence scores. We evaluate relative confidence estimation against absolute confidence estimation and self-consistency confidence methods on five state-of-the-art LMs -- GPT-4, GPT-4o, Gemini 1.5 Pro, Claude 3.5 Sonnet, and Llama 3.1 405B -- across 14 challenging STEM, social science, and commonsense reasoning question answering tasks. Our results demonstrate that relative confidence estimation consistently provides more reliable confidence scores than absolute confidence estimation, with average gains of 3.5% in selective classification AUC over direct absolute confidence estimation methods and 1.7% over self-consistency approaches across all models and datasets.

Ananya Kumar, Aditi Raghunathan, Robbie Jones et al.

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 this paper, we find that fine-tuning can achieve worse accuracy than linear probing out-of-distribution (OOD) when the pretrained features are good and the distribution shift is large. On 10 distribution shift datasets (Breeds-Living17, Breeds-Entity30, DomainNet, CIFAR $\to$ STL, CIFAR10.1, FMoW, ImageNetV2, ImageNet-R, ImageNet-A, ImageNet-Sketch), fine-tuning obtains on average 2% higher accuracy ID but 7% lower accuracy OOD than linear probing. We show theoretically that this tradeoff between ID and OOD accuracy arises even in a simple setting: fine-tuning overparameterized two-layer linear networks. We prove that the OOD error of fine-tuning is high when we initialize with a fixed or random head -- this is because while fine-tuning learns the head, the lower layers of the neural network change simultaneously and distort the pretrained features. Our analysis suggests that the easy two-step strategy of linear probing then full fine-tuning (LP-FT), sometimes used as a fine-tuning heuristic, combines the benefits of both fine-tuning and linear probing. Empirically, LP-FT outperforms both fine-tuning and linear probing on the above datasets (1% better ID, 10% better OOD than full fine-tuning).

Fahim Tajwar, Ananya Kumar, Sang Michael Xie et al.

Out-of-distribution detection is an important component of reliable ML systems. Prior literature has proposed various methods (e.g., MSP (Hendrycks & Gimpel, 2017), ODIN (Liang et al., 2018), Mahalanobis (Lee et al., 2018)), claiming they are state-of-the-art by showing they outperform previous methods on a selected set of in-distribution (ID) and out-of-distribution (OOD) datasets. In this work, we show that none of these methods are inherently better at OOD detection than others on a standardized set of 16 (ID, OOD) pairs. We give possible explanations for these inconsistencies with simple toy datasets where whether one method outperforms another depends on the structure of the ID and OOD datasets in question. Finally, we show that a method outperforming another on a certain (ID, OOD) pair may not do so in a low-data regime. In the low-data regime, we propose a distance-based method, Pairwise OOD detection (POD), which is based on Siamese networks and improves over Mahalanobis by sidestepping the expensive covariance estimation step. Our results suggest that the OOD detection problem may be too broad, and we should consider more specific structures for leverage.

Rishi Bommasani, Drew A. Hudson, Ehsan Adeli et al.

AI is undergoing a paradigm shift with the rise of models (e.g., BERT, DALL-E, GPT-3) that are trained on broad data at scale and are adaptable to a wide range of downstream tasks. We call these models foundation models to underscore their critically central yet incomplete character. This report provides a thorough account of the opportunities and risks of foundation models, ranging from their capabilities (e.g., language, vision, robotics, reasoning, human interaction) and technical principles(e.g., model architectures, training procedures, data, systems, security, evaluation, theory) to their applications (e.g., law, healthcare, education) and societal impact (e.g., inequity, misuse, economic and environmental impact, legal and ethical considerations). Though foundation models are based on standard deep learning and transfer learning, their scale results in new emergent capabilities,and their effectiveness across so many tasks incentivizes homogenization. Homogenization provides powerful leverage but demands caution, as the defects of the foundation model are inherited by all the adapted models downstream. Despite the impending widespread deployment of foundation models, we currently lack a clear understanding of how they work, when they fail, and what they are even capable of due to their emergent properties. To tackle these questions, we believe much of the critical research on foundation models will require deep interdisciplinary collaboration commensurate with their fundamentally sociotechnical nature.

Sang Michael Xie, Ananya Kumar, Robbie Jones et al.

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 input. How should we best leverage this auxiliary information for the prediction task? Empirically across three image and time-series datasets, and theoretically in a multi-task linear regression setting, we show that (i) using auxiliary information as input features improves in-distribution error but can hurt OOD error; but (ii) using auxiliary information as outputs of auxiliary pre-training tasks improves OOD error. To get the best of both worlds, we introduce In-N-Out, which first trains a model with auxiliary inputs and uses it to pseudolabel all the in-distribution inputs, then pre-trains a model on OOD auxiliary outputs and fine-tunes this model with the pseudolabels (self-training). We show both theoretically and empirically that In-N-Out outperforms auxiliary inputs or outputs alone on both in-distribution and OOD error.

Erik Jones, Shiori Sagawa, Pang Wei Koh et al.

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 simultaneously magnify existing accuracy disparities between various groups within a population, especially in the presence of spurious correlations. We observe this behavior consistently across five vision and NLP datasets. Surprisingly, increasing abstentions can even decrease accuracies on some groups. To better understand this phenomenon, we study the margin distribution, which captures the model's confidences over all predictions. For symmetric margin distributions, we prove that whether selective classification monotonically improves or worsens accuracy is fully determined by the accuracy at full coverage (i.e., without any abstentions) and whether the distribution satisfies a property we call left-log-concavity. Our analysis also shows that selective classification tends to magnify full-coverage accuracy disparities. Motivated by our analysis, we train distributionally-robust models that achieve similar full-coverage accuracies across groups and show that selective classification uniformly improves each group on these models. Altogether, our results suggest that selective classification should be used with care and underscore the importance of training models to perform equally well across groups at full coverage.

Yining Chen, Colin Wei, Ananya Kumar et al.

In unsupervised domain adaptation, existing theory focuses on situations where the source and target domains are close. In practice, conditional entropy minimization and pseudo-labeling work even when the domain shifts are much larger than those analyzed by existing theory. We identify and analyze one particular setting where the domain shift can be large, but these algorithms provably work: certain spurious features correlate with the label in the source domain but are independent of the label in the target. Our analysis considers linear classification where the spurious features are Gaussian and the non-spurious features are a mixture of log-concave distributions. For this setting, we prove that entropy minimization on unlabeled target data will avoid using the spurious feature if initialized with a decently accurate source classifier, even though the objective is non-convex and contains multiple bad local minima using the spurious features. We verify our theory for spurious domain shift tasks on semi-synthetic Celeb-A and MNIST datasets. Our results suggest that practitioners collect and self-train on large, diverse datasets to reduce biases in classifiers even if labeling is impractical.

Ananya Kumar, Tengyu Ma, Percy Liang

Machine learning systems must adapt to data distributions that evolve over time, in applications ranging from sensor networks and self-driving car perception modules to brain-machine interfaces. We consider gradual domain adaptation, where the goal is to adapt an initial classifier trained on a source domain given only unlabeled data that shifts gradually in distribution towards a target domain. We prove the first non-vacuous upper bound on the error of self-training with gradual shifts, under settings where directly adapting to the target domain can result in unbounded error. The theoretical analysis leads to algorithmic insights, highlighting that regularization and label sharpening are essential even when we have infinite data, and suggesting that self-training works particularly well for shifts with small Wasserstein-infinity distance. Leveraging the gradual shift structure leads to higher accuracies on a rotating MNIST dataset and a realistic Portraits dataset.

Ananya Kumar, Percy Liang, Tengyu Ma

Applications such as weather forecasting and personalized medicine demand models that output calibrated probability estimates---those representative of the true likelihood of a prediction. Most models are not calibrated out of the box but are recalibrated by post-processing model outputs. We find in this work that popular recalibration methods like Platt scaling and temperature scaling are (i) less calibrated than reported, and (ii) current techniques cannot estimate how miscalibrated they are. An alternative method, histogram binning, has measurable calibration error but is sample inefficient---it requires $O(B/ε^2)$ samples, compared to $O(1/ε^2)$ for scaling methods, where $B$ is the number of distinct probabilities the model can output. To get the best of both worlds, we introduce the scaling-binning calibrator, which first fits a parametric function to reduce variance and then bins the function values to actually ensure calibration. This requires only $O(1/ε^2 + B)$ samples. Next, we show that we can estimate a model's calibration error more accurately using an estimator from the meteorological community---or equivalently measure its calibration error with fewer samples ($O(\sqrt{B})$ instead of $O(B)$). We validate our approach with multiclass calibration experiments on CIFAR-10 and ImageNet, where we obtain a 35% lower calibration error than histogram binning and, unlike scaling methods, guarantees on true calibration. In these experiments, we also estimate the calibration error and ECE more accurately than the commonly used plugin estimators. We implement all these methods in a Python library: https://pypi.org/project/uncertainty-calibration

Jonathan Uesato, Ananya Kumar, Csaba Szepesvari et al.

This paper addresses the problem of evaluating learning systems in safety critical domains such as autonomous driving, where failures can have catastrophic consequences. We focus on two problems: searching for scenarios when learned agents fail and assessing their probability of failure. The standard method for agent evaluation in reinforcement learning, Vanilla Monte Carlo, can miss failures entirely, leading to the deployment of unsafe agents. We demonstrate this is an issue for current agents, where even matching the compute used for training is sometimes insufficient for evaluation. To address this shortcoming, we draw upon the rare event probability estimation literature and propose an adversarial evaluation approach. Our approach focuses evaluation on adversarially chosen situations, while still providing unbiased estimates of failure probabilities. The key difficulty is in identifying these adversarial situations -- since failures are rare there is little signal to drive optimization. To solve this we propose a continuation approach that learns failure modes in related but less robust agents. Our approach also allows reuse of data already collected for training the agent. We demonstrate the efficacy of adversarial evaluation on two standard domains: humanoid control and simulated driving. Experimental results show that our methods can find catastrophic failures and estimate failures rates of agents multiple orders of magnitude faster than standard evaluation schemes, in minutes to hours rather than days.

Ananya Kumar, S. M. Ali Eslami, Danilo J. Rezende et al.

Stochastic video prediction models take in a sequence of image frames, and generate a sequence of consecutive future image frames. These models typically generate future frames in an autoregressive fashion, which is slow and requires the input and output frames to be consecutive. We introduce a model that overcomes these drawbacks by generating a latent representation from an arbitrary set of frames that can then be used to simultaneously and efficiently sample temporally consistent frames at arbitrary time-points. For example, our model can "jump" and directly sample frames at the end of the video, without sampling intermediate frames. Synthetic video evaluations confirm substantial gains in speed and functionality without loss in fidelity. We also apply our framework to a 3D scene reconstruction dataset. Here, our model is conditioned on camera location and can sample consistent sets of images for what an occluded region of a 3D scene might look like, even if there are multiple possibilities for what that region might contain. Reconstructions and videos are available at https://bit.ly/2O4Pc4R.