Clara Fannjiang

Anastasios N. Angelopoulos, Stephen Bates, Clara Fannjiang et al. · berkeley

Prediction-powered inference is a framework for performing valid statistical inference when an experimental dataset is supplemented with predictions from a machine-learning system. The framework yields simple algorithms for computing provably valid confidence intervals for quantities such as means, quantiles, and linear and logistic regression coefficients, without making any assumptions on the machine-learning algorithm that supplies the predictions. Furthermore, more accurate predictions translate to smaller confidence intervals. Prediction-powered inference could enable researchers to draw valid and more data-efficient conclusions using machine learning. The benefits of prediction-powered inference are demonstrated with datasets from proteomics, astronomy, genomics, remote sensing, census analysis, and ecology.

Clara Fannjiang, Jennifer Listgarten

Machine learning-based design has gained traction in the sciences, most notably in the design of small molecules, materials, and proteins, with societal implications spanning drug development and manufacturing, plastic degradation, and carbon sequestration. When designing objects to achieve novel property values with machine learning, one faces a fundamental challenge: how to push past the frontier of current knowledge, distilled from the training data into the model, in a manner that rationally controls the risk of failure. If one trusts learned models too much in extrapolation, one is likely to design rubbish. In contrast, if one does not extrapolate, one cannot find novelty. Herein, we ponder how one might strike a useful balance between these two extremes. We focus in particular on designing proteins with novel property values, although much of our discussion addresses machine learning-based design more broadly.

Drew Prinster, Clara Fannjiang, Ji Won Park et al.

An agent must try new behaviors to explore and improve. In high-stakes environments, an agent that violates safety constraints may cause harm and must be taken offline, curtailing any future interaction. Imitating old behavior is safe, but excessive conservatism discourages exploration. How much behavior change is too much? We show how to use any safe reference policy as a probabilistic regulator for any optimized but untested policy. Conformal calibration on data from the safe policy determines how aggressively the new policy can act, while provably enforcing the user's declared risk tolerance. Unlike conservative optimization methods, we do not assume the user has identified the correct model class nor tuned any hyperparameters. Unlike previous conformal methods, our theory provides finite-sample guarantees even for non-monotonic bounded constraint functions. Our experiments on applications ranging from natural language question answering to biomolecular engineering show that safe exploration is not only possible from the first moment of deployment, but can also improve performance.

Shuvom Sadhuka, Drew Prinster, Clara Fannjiang et al.

Agentic AI systems execute a sequence of actions, such as reasoning steps or tool calls, in response to a user prompt. To evaluate the success of their trajectories, researchers have developed verifiers, such as LLM judges and process-reward models, to score the quality of each action in an agent's trajectory. Although these heuristic scores can be informative, there are no guarantees of correctness when used to decide whether an agent will yield a successful output. Here, we introduce e-valuator, a method to convert any black-box verifier score into a decision rule with provable control of false alarm rates. We frame the problem of distinguishing successful trajectories (that is, a sequence of actions that will lead to a correct response to the user's prompt) and unsuccessful trajectories as a sequential hypothesis testing problem. E-valuator builds on tools from e-processes to develop a sequential hypothesis test that remains statistically valid at every step of an agent's trajectory, enabling online monitoring of agents over arbitrarily long sequences of actions. Empirically, we demonstrate that e-valuator provides greater statistical power and better false alarm rate control than other strategies across six datasets and three agents. We additionally show that e-valuator can be used for to quickly terminate problematic trajectories and save tokens. Together, e-valuator provides a lightweight, model-agnostic framework that converts verifier heuristics into decisions rules with statistical guarantees, enabling the deployment of more reliable agentic systems.

Clara Fannjiang, Ji Won Park

Algorithms for machine learning-guided design, or design algorithms, use machine learning-based predictions to propose novel objects with desired property values. Given a new design task -- for example, to design novel proteins with high binding affinity to a therapeutic target -- one must choose a design algorithm and specify any hyperparameters and predictive and/or generative models involved. How can these decisions be made such that the resulting designs are successful? This paper proposes a method for design algorithm selection, which aims to select design algorithms that will produce a distribution of design labels satisfying a user-specified success criterion -- for example, that at least ten percent of designs' labels exceed a threshold. It does so by combining designs' predicted property values with held-out labeled data to reliably forecast characteristics of the label distributions produced by different design algorithms, building upon techniques from prediction-powered inference. The method is guaranteed with high probability to return design algorithms that yield successful label distributions (or the null set if none exist), if the density ratios between the design and labeled data distributions are known. We demonstrate the method's effectiveness in simulated protein and RNA design tasks, in settings with either known or estimated density ratios.

Clara Fannjiang, Stephen Bates, Anastasios N. Angelopoulos et al.

Many applications of machine learning methods involve an iterative protocol in which data are collected, a model is trained, and then outputs of that model are used to choose what data to consider next. For example, one data-driven approach for designing proteins is to train a regression model to predict the fitness of protein sequences, then use it to propose new sequences believed to exhibit greater fitness than observed in the training data. Since validating designed sequences in the wet lab is typically costly, it is important to quantify the uncertainty in the model's predictions. This is challenging because of a characteristic type of distribution shift between the training and test data in the design setting -- one in which the training and test data are statistically dependent, as the latter is chosen based on the former. Consequently, the model's error on the test data -- that is, the designed sequences -- has an unknown and possibly complex relationship with its error on the training data. We introduce a method to quantify predictive uncertainty in such settings. We do so by constructing confidence sets for predictions that account for the dependence between the training and test data. The confidence sets we construct have finite-sample guarantees that hold for any prediction algorithm, even when a trained model chooses the test-time input distribution. As a motivating use case, we demonstrate with several real data sets how our method quantifies uncertainty for the predicted fitness of designed proteins, and can therefore be used to select design algorithms that achieve acceptable trade-offs between high predicted fitness and low predictive uncertainty.

Ghassen Jerfel, Serena Wang, Clara Fannjiang et al.

Variational Inference (VI) is a popular alternative to asymptotically exact sampling in Bayesian inference. Its main workhorse is optimization over a reverse Kullback-Leibler divergence (RKL), which typically underestimates the tail of the posterior leading to miscalibration and potential degeneracy. Importance sampling (IS), on the other hand, is often used to fine-tune and de-bias the estimates of approximate Bayesian inference procedures. The quality of IS crucially depends on the choice of the proposal distribution. Ideally, the proposal distribution has heavier tails than the target, which is rarely achievable by minimizing the RKL. We thus propose a novel combination of optimization and sampling techniques for approximate Bayesian inference by constructing an IS proposal distribution through the minimization of a forward KL (FKL) divergence. This approach guarantees asymptotic consistency and a fast convergence towards both the optimal IS estimator and the optimal variational approximation. We empirically demonstrate on real data that our method is competitive with variational boosting and MCMC.

Clara Fannjiang, Jennifer Listgarten

Data-driven design is making headway into a number of application areas, including protein, small-molecule, and materials engineering. The design goal is to construct an object with desired properties, such as a protein that binds to a therapeutic target, or a superconducting material with a higher critical temperature than previously observed. To that end, costly experimental measurements are being replaced with calls to high-capacity regression models trained on labeled data, which can be leveraged in an in silico search for design candidates. However, the design goal necessitates moving into regions of the design space beyond where such models were trained. Therefore, one can ask: should the regression model be altered as the design algorithm explores the design space, in the absence of new data? Herein, we answer this question in the affirmative. In particular, we (i) formalize the data-driven design problem as a non-zero-sum game, (ii) develop a principled strategy for retraining the regression model as the design algorithm proceeds---what we refer to as autofocusing, and (iii) demonstrate the promise of autofocusing empirically.

David H. Brookes, Akosua Busia, Clara Fannjiang et al.

We show that a large class of Estimation of Distribution Algorithms, including, but not limited to, Covariance Matrix Adaption, can be written as a Monte Carlo Expectation-Maximization algorithm, and as exact EM in the limit of infinite samples. Because EM sits on a rigorous statistical foundation and has been thoroughly analyzed, this connection provides a new coherent framework with which to reason about EDAs.