arXiv subjects

Anders Andreassen

Publications and source records attributed to Anders Andreassen.

At least 19 recordsLinked to original sources

arXiv2025

Gemini: A Family of Highly Capable Multimodal Models

This report introduces a new family of multimodal models, Gemini, that exhibit remarkable capabilities across image, audio, video, and text understanding. The Gemini family consists of Ultra, Pro, and Nano sizes, suitable for applications ranging from complex reasoning tasks to on-device memory-constrained use-cases. Evaluation on a broad range of benchmarks shows that our most-capable Gemini Ultra model advances the state of the art in 30 of 32 of these benchmarks - notably being the first model to achieve human-expert performance on the well-studied exam benchmark MMLU, and improving the state of the art in every one of the 20 multimodal benchmarks we examined. We believe that the new capabilities of the Gemini family in cross-modal reasoning and language understanding will enable a wide variety of use cases. We discuss our approach toward post-training and deploying Gemini models responsibly to users through services including Gemini, Gemini Advanced, Google AI Studio, and Cloud Vertex AI.

cs.CL↗

arXiv2024

Understanding the Failure Modes of Out-of-Distribution Generalization

Empirical studies suggest that machine learning models often rely on features, such as the background, that may be spuriously correlated with the label only during training time, resulting in poor accuracy during test-time. In this work, we identify the fundamental factors that give rise to this behavior, by explaining why models fail this way {\em even} in easy-to-learn tasks where one would expect these models to succeed. In particular, through a theoretical study of gradient-descent-trained linear classifiers on some easy-to-learn tasks, we uncover two complementary failure modes. These modes arise from how spurious correlations induce two kinds of skews in the data: one geometric in nature, and another, statistical in nature. Finally, we construct natural modifications of image classification datasets to understand when these failure modes can arise in practice. We also design experiments to isolate the two failure modes when training modern neural networks on these datasets.

cs.LG↗

arXiv2023

Beyond the Imitation Game: Quantifying and extrapolating the capabilities of language models

Language models demonstrate both quantitative improvement and new qualitative capabilities with increasing scale. Despite their potentially transformative impact, these new capabilities are as yet poorly characterized. In order to inform future research, prepare for disruptive new model capabilities, and ameliorate socially harmful effects, it is vital that we understand the present and near-future capabilities and limitations of language models. To address this challenge, we introduce the Beyond the Imitation Game benchmark (BIG-bench). BIG-bench currently consists of 204 tasks, contributed by 450 authors across 132 institutions. Task topics are diverse, drawing problems from linguistics, childhood development, math, common-sense reasoning, biology, physics, social bias, software development, and beyond. BIG-bench focuses on tasks that are believed to be beyond the capabilities of current language models. We evaluate the behavior of OpenAI's GPT models, Google-internal dense transformer architectures, and Switch-style sparse transformers on BIG-bench, across model sizes spanning millions to hundreds of billions of parameters. In addition, a team of human expert raters performed all tasks in order to provide a strong baseline. Findings include: model performance and calibration both improve with scale, but are poor in absolute terms (and when compared with rater performance); performance is remarkably similar across model classes, though with benefits from sparsity; tasks that improve gradually and predictably commonly involve a large knowledge or memorization component, whereas tasks that exhibit "breakthrough" behavior at a critical scale often involve multiple steps or components, or brittle metrics; social bias typically increases with scale in settings with ambiguous context, but this can be improved with prompting.

cs.CL↗

arXiv2022

Exploring Length Generalization in Large Language Models

The ability to extrapolate from short problem instances to longer ones is an important form of out-of-distribution generalization in reasoning tasks, and is crucial when learning from datasets where longer problem instances are rare. These include theorem proving, solving quantitative mathematics problems, and reading/summarizing novels. In this paper, we run careful empirical studies exploring the length generalization capabilities of transformer-based language models. We first establish that naively finetuning transformers on length generalization tasks shows significant generalization deficiencies independent of model scale. We then show that combining pretrained large language models' in-context learning abilities with scratchpad prompting (asking the model to output solution steps before producing an answer) results in a dramatic improvement in length generalization. We run careful failure analyses on each of the learning modalities and identify common sources of mistakes that highlight opportunities in equipping language models with the ability to generalize to longer problems.

cs.CL↗

arXiv2022

Solving Quantitative Reasoning Problems with Language Models

Language models have achieved remarkable performance on a wide range of tasks that require natural language understanding. Nevertheless, state-of-the-art models have generally struggled with tasks that require quantitative reasoning, such as solving mathematics, science, and engineering problems at the college level. To help close this gap, we introduce Minerva, a large language model pretrained on general natural language data and further trained on technical content. The model achieves state-of-the-art performance on technical benchmarks without the use of external tools. We also evaluate our model on over two hundred undergraduate-level problems in physics, biology, chemistry, economics, and other sciences that require quantitative reasoning, and find that the model can correctly answer nearly a third of them.

cs.CL↗

arXiv2021

The Evolution of Out-of-Distribution Robustness Throughout Fine-Tuning

Although machine learning models typically experience a drop in performance on out-of-distribution data, accuracies on in- versus out-of-distribution data are widely observed to follow a single linear trend when evaluated across a testbed of models. Models that are more accurate on the out-of-distribution data relative to this baseline exhibit "effective robustness" and are exceedingly rare. Identifying such models, and understanding their properties, is key to improving out-of-distribution performance. We conduct a thorough empirical investigation of effective robustness during fine-tuning and surprisingly find that models pre-trained on larger datasets exhibit effective robustness during training that vanishes at convergence. We study how properties of the data influence effective robustness, and we show that it increases with the larger size, more diversity, and higher example difficulty of the dataset. We also find that models that display effective robustness are able to correctly classify 10% of the examples that no other current testbed model gets correct. Finally, we discuss several strategies for scaling effective robustness to the high-accuracy regime to improve the out-of-distribution accuracy of state-of-the-art models.

cs.LG↗

arXiv2021

Scaffolding Simulations with Deep Learning for High-dimensional Deconvolution

A common setting for scientific inference is the ability to sample from a high-fidelity forward model (simulation) without having an explicit probability density of the data. We propose a simulation-based maximum likelihood deconvolution approach in this setting called OmniFold. Deep learning enables this approach to be naturally unbinned and (variable-, and) high-dimensional. In contrast to model parameter estimation, the goal of deconvolution is to remove detector distortions in order to enable a variety of down-stream inference tasks. Our approach is the deep learning generalization of the common Richardson-Lucy approach that is also called Iterative Bayesian Unfolding in particle physics. We show how OmniFold can not only remove detector distortions, but it can also account for noise processes and acceptance effects.

stat.ML↗

arXiv2021

Parameter Estimation using Neural Networks in the Presence of Detector Effects

Histogram-based template fits are the main technique used for estimating parameters of high energy physics Monte Carlo generators. Parametrized neural network reweighting can be used to extend this fitting procedure to many dimensions and does not require binning. If the fit is to be performed using reconstructed data, then expensive detector simulations must be used for training the neural networks. We introduce a new two-level fitting approach that only requires one dataset with detector simulation and then a set of additional generation-level datasets without detector effects included. This Simulation-level fit based on Reweighting Generator-level events with Neural networks (SRGN) is demonstrated using simulated datasets for a variety of examples including a simple Gaussian random variable, parton shower tuning, and the top quark mass extraction.

hep-ph↗

arXiv2021

The LHC Olympics 2020: A Community Challenge for Anomaly Detection in High Energy Physics

A new paradigm for data-driven, model-agnostic new physics searches at colliders is emerging, and aims to leverage recent breakthroughs in anomaly detection and machine learning. In order to develop and benchmark new anomaly detection methods within this framework, it is essential to have standard datasets. To this end, we have created the LHC Olympics 2020, a community challenge accompanied by a set of simulated collider events. Participants in these Olympics have developed their methods using an R&D dataset and then tested them on black boxes: datasets with an unknown anomaly (or not). This paper will review the LHC Olympics 2020 challenge, including an overview of the competition, a description of methods deployed in the competition, lessons learned from the experience, and implications for data analyses with future datasets as well as future colliders.

hep-ph↗

arXiv2020

Asymptotics of Wide Convolutional Neural Networks

Wide neural networks have proven to be a rich class of architectures for both theory and practice. Motivated by the observation that finite width convolutional networks appear to outperform infinite width networks, we study scaling laws for wide CNNs and networks with skip connections. Following the approach of (Dyer & Gur-Ari, 2019), we present a simple diagrammatic recipe to derive the asymptotic width dependence for many quantities of interest. These scaling relationships provide a solvable description for the training dynamics of wide convolutional networks. We test these relations across a broad range of architectures. In particular, we find that the difference in performance between finite and infinite width models vanishes at a definite rate with respect to model width. Nonetheless, this relation is consistent with finite width models generalizing either better or worse than their infinite width counterparts, and we provide examples where the relative performance depends on the optimization details.

cs.LG↗

arXiv2020

OmniFold: A Method to Simultaneously Unfold All Observables

Collider data must be corrected for detector effects ("unfolded") to be compared with many theoretical calculations and measurements from other experiments. Unfolding is traditionally done for individual, binned observables without including all information relevant for characterizing the detector response. We introduce OmniFold, an unfolding method that iteratively reweights a simulated dataset, using machine learning to capitalize on all available information. Our approach is unbinned, works for arbitrarily high-dimensional data, and naturally incorporates information from the full phase space. We illustrate this technique on a realistic jet substructure example from the Large Hadron Collider and compare it to standard binned unfolding methods. This new paradigm enables the simultaneous measurement of all observables, including those not yet invented at the time of the analysis.

hep-ph↗

arXiv2020

Simulation Assisted Likelihood-free Anomaly Detection

Given the lack of evidence for new particle discoveries at the Large Hadron Collider (LHC), it is critical to broaden the search program. A variety of model-independent searches have been proposed, adding sensitivity to unexpected signals. There are generally two types of such searches: those that rely heavily on simulations and those that are entirely based on (unlabeled) data. This paper introduces a hybrid method that makes the best of both approaches. For potential signals that are resonant in one known feature, this new method first learns a parameterized reweighting function to morph a given simulation to match the data in sidebands. This function is then interpolated into the signal region and then the reweighted background-only simulation can be used for supervised learning as well as for background estimation. The background estimation from the reweighted simulation allows for non-trivial correlations between features used for classification and the resonant feature. A dijet search with jet substructure is used to illustrate the new method. Future applications of Simulation Assisted Likelihood-free Anomaly Detection (SALAD) include a variety of final states and potential combinations with other model-independent approaches.

hep-ph↗

arXiv2019

Neural Networks for Full Phase-space Reweighting and Parameter Tuning

Precise scientific analysis in collider-based particle physics is possible because of complex simulations that connect fundamental theories to observable quantities. The significant computational cost of these programs limits the scope, precision, and accuracy of Standard Model measurements and searches for new phenomena. We therefore introduce Deep neural networks using Classification for Tuning and Reweighting (DCTR), a neural network-based approach to reweight and fit simulations using all kinematic and flavor information -- the full phase space. DCTR can perform tasks that are currently not possible with existing methods, such as estimating non-perturbative fragmentation uncertainties. The core idea behind the new approach is to exploit powerful high-dimensional classifiers to reweight phase space as well as to identify the best parameters for describing data. Numerical examples from $e^+e^-\rightarrow\text{jets}$ demonstrate the fidelity of these methods for simulation parameters that have a big and broad impact on phase space as well as those that have a minimal and/or localized impact. The high fidelity of the full phase-space reweighting enables a new paradigm for simulations, parameter tuning, and model systematic uncertainties across particle physics and possibly beyond.

hep-ph↗

arXiv2019

Binary JUNIPR: an interpretable probabilistic model for discrimination

JUNIPR is an approach to unsupervised learning in particle physics that scaffolds a probabilistic model for jets around their representation as binary trees. Separate JUNIPR models can be learned for different event or jet types, then compared and explored for physical insight. The relative probabilities can also be used for discrimination. In this paper, we show how the training of the separate models can be refined in the context of classification to optimize discrimination power. We refer to this refined approach as Binary JUNIPR. Binary JUNIPR achieves state-of-the-art performance for quark/gluon discrimination and top-tagging. The trained models can then be analyzed to provide physical insight into how the classification is achieved. As examples, we explore differences between quark and gluon jets and between gluon jets generated with two different simulations.

hep-ph↗

arXiv2018

Scale Invariant Instantons and the Complete Lifetime of the Standard Model

In a classically scale-invariant quantum field theory, tunneling rates are infrared divergent due to the existence of instantons of any size. While one expects such divergences to be resolved by quantum effects, it has been unclear how higher-loop corrections can resolve a problem appearing already at one loop. With a careful power counting, we uncover a series of loop contributions that dominate over the one-loop result and sum all the necessary terms. We also clarify previously incomplete treatments of related issues pertaining to global symmetries, gauge fixing and finite mass effects. In addition, we produce exact closed-form solutions for the functional determinants over scalars, fermions and vector bosons around the scale-invariant bounce, demonstrating manifest gauge invariance in the vector case. With these problems solved, we produce the first complete calculation of the lifetime of our universe: 10^139 years. With 95% confidence, we expect our universe to last more than 10^58 years. The uncertainty is part experimental uncertainty on the top quark mass and on $αs$ and part theory uncertainty from electroweak threshold corrections. Using our complete result, we provide phase diagrams in the $mt/mh$ and the $mt/αs$ planes, with uncertainty bands. To rule out absolute stability to $3σ$ confidence, the uncertainty on the top quark pole mass would have to be pushed below 250 MeV or the uncertainty on $αs(mZ)$ pushed below 0.00025.

hep-ph↗

arXiv2018

JUNIPR: a Framework for Unsupervised Machine Learning in Particle Physics

In applications of machine learning to particle physics, a persistent challenge is how to go beyond discrimination to learn about the underlying physics. To this end, a powerful tool would be a framework for unsupervised learning, where the machine learns the intricate high-dimensional contours of the data upon which it is trained, without reference to pre-established labels. In order to approach such a complex task, an unsupervised network must be structured intelligently, based on a qualitative understanding of the data. In this paper, we scaffold the neural network's architecture around a leading-order model of the physics underlying the data. In addition to making unsupervised learning tractable, this design actually alleviates existing tensions between performance and interpretability. We call the framework JUNIPR: "Jets from UNsupervised Interpretable PRobabilistic models". In this approach, the set of particle momenta composing a jet are clustered into a binary tree that the neural network examines sequentially. Training is unsupervised and unrestricted: the network could decide that the data bears little correspondence to the chosen tree structure. However, when there is a correspondence, the network's output along the tree has a direct physical interpretation. JUNIPR models can perform discrimination tasks, through the statistically optimal likelihood-ratio test, and they permit visualizations of discrimination power at each branching in a jet's tree. Additionally, JUNIPR models provide a probability distribution from which events can be drawn, providing a data-driven Monte Carlo generator. As a third application, JUNIPR models can reweight events from one (e.g. simulated) data set to agree with distributions from another (e.g. experimental) data set.

hep-ph↗

arXiv2017

Precision decay rate calculations in quantum field theory

Tunneling in quantum field theory is worth understanding properly, not least because it controls the long term fate of our universe. There are however, a number of features of tunneling rate calculations which lack a desirable transparency, such as the necessity of analytic continuation, the appropriateness of using an effective instead of classical potential, and the sensitivity to short-distance physics. This paper attempts to review in pedagogical detail the physical origin of tunneling and its connection to the path integral. Both the traditional potential-deformation method and a recent more direct propagator-based method are discussed. Some new insights from using approximate semi-classical solutions are presented. In addition, we explore the sensitivity of the lifetime of our universe to short distance physics, such as quantum gravity, emphasizing a number of important subtleties.

hep-th↗

arXiv2017

Reducing the Top Quark Mass Uncertainty with Jet Grooming

The measurement of the top quark mass has large systematic uncertainties coming from the Monte Carlo simulations that are used to match theory and experiment. We explore how much that uncertainty can be reduced by using jet grooming procedures. We estimate the inherent ambiguity in what is meant by Monte Carlo mass to be around 530 MeV without any corrections. This uncertainty can be reduced by 60% to 200 MeV by calibrating to the W mass and a further 33% to 140 MeV by applying soft-drop jet grooming (or by 20% more to 170 MeV with trimming). At e+e- colliders, the associated uncertainty is around 110 MeV, reducing to 50 MeV after calibrating to the W mass. By analyzing the tuning parameters, we conclude that the importance of jet grooming after calibrating to the W mass is to reduce sensitivity to the underlying event.

hep-ph↗