Uncovering mesa-optimization algorithms in Transformers

Transformers have become the dominant model in deep learning, but the reason for their superior performance is poorly understood. Here, we hypothesize that the strong performance of Transformers stems from an architectural bias towards mesa-optimization, a learned process running within the forward pass of a model consisting of the following two steps: (i) the construction of an internal learning objective, and (ii) its corresponding solution found through optimization. To test this hypothesis, we reverse-engineer a series of autoregressive Transformers trained on simple sequence modeling tasks, uncovering underlying gradient-based mesa-optimization algorithms driving the generation of predictions. Moreover, we show that the learned forward-pass optimization algorithm can be immediately repurposed to solve supervised few-shot tasks, suggesting that mesa-optimization might underlie the in-context learning capabilities of large language models. Finally, we propose a novel self-attention layer, the mesa-layer, that explicitly and efficiently solves optimization problems specified in context. We find that this layer can lead to improved performance in synthetic and preliminary language modeling experiments, adding weight to our hypothesis that mesa-optimization is an important operation hidden within the weights of trained Transformers.

Evaluating the Moral Beliefs Encoded in LLMs

This paper presents a case study on the design, administration, post-processing, and evaluation of surveys on large language models (LLMs). It comprises two components: (1) A statistical method for eliciting beliefs encoded in LLMs. We introduce statistical measures and evaluation metrics that quantify the probability of an LLM "making a choice", the associated uncertainty, and the consistency of that choice. (2) We apply this method to study what moral beliefs are encoded in different LLMs, especially in ambiguous cases where the right choice is not obvious. We design a large-scale survey comprising 680 high-ambiguity moral scenarios (e.g., "Should I tell a white lie?") and 687 low-ambiguity moral scenarios (e.g., "Should I stop for a pedestrian on the road?"). Each scenario includes a description, two possible actions, and auxiliary labels indicating violated rules (e.g., "do not kill"). We administer the survey to 28 open- and closed-source LLMs. We find that (a) in unambiguous scenarios, most models "choose" actions that align with commonsense. In ambiguous cases, most models express uncertainty. (b) Some models are uncertain about choosing the commonsense action because their responses are sensitive to the question-wording. (c) Some models reflect clear preferences in ambiguous scenarios. Specifically, closed-source models tend to agree with each other.

Trust Your $\nabla$: Gradient-based Intervention Targeting for Causal Discovery

Inferring causal structure from data is a challenging task of fundamental importance in science. Observational data are often insufficient to identify a system's causal structure uniquely. While conducting interventions (i.e., experiments) can improve the identifiability, such samples are usually challenging and expensive to obtain. Hence, experimental design approaches for causal discovery aim to minimize the number of interventions by estimating the most informative intervention target. In this work, we propose a novel Gradient-based Intervention Targeting method, abbreviated GIT, that 'trusts' the gradient estimator of a gradient-based causal discovery framework to provide signals for the intervention acquisition function. We provide extensive experiments in simulated and real-world datasets and demonstrate that GIT performs on par with competitive baselines, surpassing them in the low-data regime.

FED-CD: Federated Causal Discovery from Interventional and Observational Data

Causal discovery, the inference of causal relations from data, is a core task of fundamental importance in all scientific domains, and several new machine learning methods for addressing the causal discovery problem have been proposed recently. However, existing machine learning methods for causal discovery typically require that the data used for inference is pooled and available in a centralized location. In many domains of high practical importance, such as in healthcare, data is only available at local data-generating entities (e.g. hospitals in the healthcare context), and cannot be shared across entities due to, among others, privacy and regulatory reasons. In this work, we address the problem of inferring causal structure - in the form of a directed acyclic graph (DAG) - from a distributed data set that contains both observational and interventional data in a privacy-preserving manner by exchanging updates instead of samples. To this end, we introduce a new federated framework, FED-CD, that enables the discovery of global causal structures both when the set of intervened covariates is the same across decentralized entities, and when the set of intervened covariates are potentially disjoint. We perform a comprehensive experimental evaluation on synthetic data that demonstrates that FED-CD enables effective aggregation of decentralized data for causal discovery without direct sample sharing, even when the contributing distributed data sets cover disjoint sets of interventions. Effective methods for causal discovery in distributed data sets could significantly advance scientific discovery and knowledge sharing in important settings, for instance, healthcare, in which sharing of data across local sites is difficult or prohibited.

On the Generalization and Adaption Performance of Causal Models

Learning models that offer robust out-of-distribution generalization and fast adaptation is a key challenge in modern machine learning. Modelling causal structure into neural networks holds the promise to accomplish robust zero and few-shot adaptation. Recent advances in differentiable causal discovery have proposed to factorize the data generating process into a set of modules, i.e. one module for the conditional distribution of every variable where only causal parents are used as predictors. Such a modular decomposition of knowledge enables adaptation to distributions shifts by only updating a subset of parameters. In this work, we systematically study the generalization and adaption performance of such modular neural causal models by comparing it to monolithic models and structured models where the set of predictors is not constrained to causal parents. Our analysis shows that the modular neural causal models outperform other models on both zero and few-shot adaptation in low data regimes and offer robust generalization. We also found that the effects are more significant for sparser graphs as compared to denser graphs.

Learning Neural Causal Models with Active Interventions

Discovering causal structures from data is a challenging inference problem of fundamental importance in all areas of science. The appealing scaling properties of neural networks have recently led to a surge of interest in differentiable neural network-based methods for learning causal structures from data. So far differentiable causal discovery has focused on static datasets of observational or interventional origin. In this work, we introduce an active intervention-targeting mechanism which enables a quick identification of the underlying causal structure of the data-generating process. Our method significantly reduces the required number of interactions compared with random intervention targeting and is applicable for both discrete and continuous optimization formulations of learning the underlying directed acyclic graph (DAG) from data. We examine the proposed method across a wide range of settings and demonstrate superior performance on multiple benchmarks from simulated to real-world data.

Variational Causal Networks: Approximate Bayesian Inference over Causal Structures

Learning the causal structure that underlies data is a crucial step towards robust real-world decision making. The majority of existing work in causal inference focuses on determining a single directed acyclic graph (DAG) or a Markov equivalence class thereof. However, a crucial aspect to acting intelligently upon the knowledge about causal structure which has been inferred from finite data demands reasoning about its uncertainty. For instance, planning interventions to find out more about the causal mechanisms that govern our data requires quantifying epistemic uncertainty over DAGs. While Bayesian causal inference allows to do so, the posterior over DAGs becomes intractable even for a small number of variables. Aiming to overcome this issue, we propose a form of variational inference over the graphs of Structural Causal Models (SCMs). To this end, we introduce a parametric variational family modelled by an autoregressive distribution over the space of discrete DAGs. Its number of parameters does not grow exponentially with the number of variables and can be tractably learned by maximising an Evidence Lower Bound (ELBO). In our experiments, we demonstrate that the proposed variational posterior is able to provide a good approximation of the true posterior.

Improved Segmentation and Detection Sensitivity of Diffusion-Weighted Brain Infarct Lesions with Synthetically Enhanced Deep Learning

Purpose: To compare the segmentation and detection performance of a deep learning model trained on a database of human-labelled clinical diffusion-weighted (DW) stroke lesions to a model trained on the same database enhanced with synthetic DW stroke lesions. Methods: In this institutional review board approved study, a stroke database of 962 cases (mean age 65+/-17 years, 255 males, 449 scans with DW positive stroke lesions) and a normal database of 2,027 patients (mean age 38+/-24 years,1088 females) were obtained. Brain volumes with synthetic DW stroke lesions were produced by warping the relative signal increase of real strokes to normal brain volumes. A generic 3D U-Net was trained on four different databases to generate four different models: (a) 375 neuroradiologist-labeled clinical DW positive stroke cases(CDB);(b) 2,000 synthetic cases(S2DB);(c) CDB+2,000 synthetic cases(CS2DB); or (d) CDB+40,000 synthetic cases(CS40DB). The models were tested on 20%(n=192) of the cases of the stroke database, which were excluded from the training set. Segmentation accuracy was characterized using Dice score and lesion volume of the stroke segmentation, and statistical significance was tested using a paired, two-tailed, Student's t-test. Detection sensitivity and specificity was compared to three neuroradiologists. Results: The performance of the 3D U-Net model trained on the CS40DB(mean Dice 0.72) was better than models trained on the CS2DB (0.70,P <0.001) or the CDB(0.65,P<0.001). The deep learning model was also more sensitive (91%[89%-93%]) than each of the three human readers(84%[81%-87%],78%[75%-81%],and 79%[76%-82%]), but less specific(75%[72%-78%] vs for the three human readers (96%[94%-97%],92%[90%-94%] and 89%[86%-91%]). Conclusion: Deep learning training for segmentation and detection of DW stroke lesions was significantly improved by enhancing the training set with synthetic lesions.