Overparameterization is shown to result in poor test accuracy on rare subgroups under a variety of settings where subgroup information is known. To gain a more complete picture, we consider the case where subgroup information is unknown. We investigate the effect of model size on worst-group generalization under empirical risk minimization (ERM) across a wide range of settings, varying: 1) architectures (ResNet, VGG, or BERT), 2) domains (vision or natural language processing), 3) model size (width or depth), and 4) initialization (with pre-trained or random weights). Our systematic evaluation reveals that increasing model size does not hurt, and may help, worst-group test performance under ERM across all setups. In particular, increasing pre-trained model size consistently improves performance on Waterbirds and MultiNLI. We advise practitioners to use larger pre-trained models when subgroup labels are unknown.
When making everyday decisions, people are guided by their conscience, an internal sense of right and wrong. By contrast, artificial agents are currently not endowed with a moral sense. As a consequence, they may learn to behave immorally when trained on environments that ignore moral concerns, such as violent video games. With the advent of generally capable agents that pretrain on many environments, it will become necessary to mitigate inherited biases from environments that teach immoral behavior. To facilitate the development of agents that avoid causing wanton harm, we introduce Jiminy Cricket, an environment suite of 25 text-based adventure games with thousands of diverse, morally salient scenarios. By annotating every possible game state, the Jiminy Cricket environments robustly evaluate whether agents can act morally while maximizing reward. Using models with commonsense moral knowledge, we create an elementary artificial conscience that assesses and guides agents. In extensive experiments, we find that the artificial conscience approach can steer agents towards moral behavior without sacrificing performance.
Machine learning (ML) systems are rapidly increasing in size, are acquiring new capabilities, and are increasingly deployed in high-stakes settings. As with other powerful technologies, safety for ML should be a leading research priority. In response to emerging safety challenges in ML, such as those introduced by recent large-scale models, we provide a new roadmap for ML Safety and refine the technical problems that the field needs to address. We present four problems ready for research, namely withstanding hazards ("Robustness"), identifying hazards ("Monitoring"), steering ML systems ("Alignment"), and reducing risks to how ML systems are handled ("External Safety"). Throughout, we clarify each problem's motivation and provide concrete research directions.
Large-scale, two-sided matching platforms must find market outcomes that align with user preferences while simultaneously learning these preferences from data. However, since preferences are inherently uncertain during learning, the classical notion of stability (Gale and Shapley, 1962; Shapley and Shubik, 1971) is unattainable in these settings. To bridge this gap, we develop a framework and algorithms for learning stable market outcomes under uncertainty. Our primary setting is matching with transferable utilities, where the platform both matches agents and sets monetary transfers between them. We design an incentive-aware learning objective that captures the distance of a market outcome from equilibrium. Using this objective, we analyze the complexity of learning as a function of preference structure, casting learning as a stochastic multi-armed bandit problem. Algorithmically, we show that "optimism in the face of uncertainty," the principle underlying many bandit algorithms, applies to a primal-dual formulation of matching with transfers and leads to near-optimal regret bounds. Our work takes a first step toward elucidating when and how stable matchings arise in large, data-driven marketplaces.
To understand neural network behavior, recent works quantitatively compare different networks' learned representations using canonical correlation analysis (CCA), centered kernel alignment (CKA), and other dissimilarity measures. Unfortunately, these widely used measures often disagree on fundamental observations, such as whether deep networks differing only in random initialization learn similar representations. These disagreements raise the question: which, if any, of these dissimilarity measures should we believe? We provide a framework to ground this question through a concrete test: measures should have sensitivity to changes that affect functional behavior, and specificity against changes that do not. We quantify this through a variety of functional behaviors including probing accuracy and robustness to distribution shift, and examine changes such as varying random initialization and deleting principal components. We find that current metrics exhibit different weaknesses, note that a classical baseline performs surprisingly well, and highlight settings where all metrics appear to fail, thus providing a challenge set for further improvement.
While programming is one of the most broadly applicable skills in modern society, modern machine learning models still cannot code solutions to basic problems. Despite its importance, there has been surprisingly little work on evaluating code generation, and it can be difficult to accurately assess code generation performance rigorously. To meet this challenge, we introduce APPS, a benchmark for code generation. Unlike prior work in more restricted settings, our benchmark measures the ability of models to take an arbitrary natural language specification and generate satisfactory Python code. Similar to how companies assess candidate software developers, we then evaluate models by checking their generated code on test cases. Our benchmark includes 10,000 problems, which range from having simple one-line solutions to being substantial algorithmic challenges. We fine-tune large language models on both GitHub and our training set, and we find that the prevalence of syntax errors is decreasing exponentially as models improve. Recent models such as GPT-Neo can pass approximately 20% of the test cases of introductory problems, so we find that machine learning models are now beginning to learn how to code. As the social significance of automatic code generation increases over the coming years, our benchmark can provide an important measure for tracking advancements.
Larger language models have higher accuracy on average, but are they better on every single instance (datapoint)? Some work suggests larger models have higher out-of-distribution robustness, while other work suggests they have lower accuracy on rare subgroups. To understand these differences, we investigate these models at the level of individual instances. However, one major challenge is that individual predictions are highly sensitive to noise in the randomness in training. We develop statistically rigorous methods to address this, and after accounting for pretraining and finetuning noise, we find that our BERT-Large is worse than BERT-Mini on at least 1-4% of instances across MNLI, SST-2, and QQP, compared to the overall accuracy improvement of 2-10%. We also find that finetuning noise increases with model size and that instance-level accuracy has momentum: improvement from BERT-Mini to BERT-Medium correlates with improvement from BERT-Medium to BERT-Large. Our findings suggest that instance-level predictions provide a rich source of information; we therefore, recommend that researchers supplement model weights with model predictions.
Traditional learning approaches for classification implicitly assume that each mistake has the same cost. In many real-world problems though, the utility of a decision depends on the underlying context $x$ and decision $y$. However, directly incorporating these utilities into the learning objective is often infeasible since these can be quite complex and difficult for humans to specify. We formally study this as agnostic learning with unknown utilities: given a dataset $S = \{x_1, \ldots, x_n\}$ where each data point $x_i \sim \mathcal{D}$, the objective of the learner is to output a function $f$ in some class of decision functions $\mathcal{F}$ with small excess risk. This risk measures the performance of the output predictor $f$ with respect to the best predictor in the class $\mathcal{F}$ on the unknown underlying utility $u^*$. This utility $u^*$ is not assumed to have any specific structure. This raises an interesting question whether learning is even possible in our setup, given that obtaining a generalizable estimate of utility $u^*$ might not be possible from finitely many samples. Surprisingly, we show that estimating the utilities of only the sampled points~$S$ suffices to learn a decision function which generalizes well. We study mechanisms for eliciting information which allow a learner to estimate the utilities $u^*$ on the set $S$. We introduce a family of elicitation mechanisms by generalizing comparisons, called the $k$-comparison oracle, which enables the learner to ask for comparisons across $k$ different inputs $x$ at once. We show that the excess risk in our agnostic learning framework decreases at a rate of $O\left(\frac{1}{k} \right)$. This result brings out an interesting accuracy-elicitation trade-off -- as the order $k$ of the oracle increases, the comparative queries become harder to elicit from humans but allow for more accurate learning.
Adversarially trained models exhibit a large generalization gap: they can interpolate the training set even for large perturbation radii, but at the cost of large test error on clean samples. To investigate this gap, we decompose the test risk into its bias and variance components. We find that the bias increases monotonically with perturbation size and is the dominant term in the risk. Meanwhile, the variance is unimodal, peaking near the interpolation threshold for the training set. In contrast, we show that popular explanations for the generalization gap instead predict the variance to be monotonic, which leaves an unresolved mystery. We show that the same unimodal variance appears in a simple high-dimensional logistic regression problem, as well as for randomized smoothing. Overall, our results highlight the power of bias-variance decompositions in modern settings--by providing two measurements instead of one, they can rule out some theories and clarify others.