Silvia Sapora

Sascha Frey, Kang Li, Peer Nagy et al.

Financial exchanges across the world use limit order books (LOBs) to process orders and match trades. For research purposes it is important to have large scale efficient simulators of LOB dynamics. LOB simulators have previously been implemented in the context of agent-based models (ABMs), reinforcement learning (RL) environments, and generative models, processing order flows from historical data sets and hand-crafted agents alike. For many applications, there is a requirement for processing multiple books, either for the calibration of ABMs or for the training of RL agents. We showcase the first GPU-enabled LOB simulator designed to process thousands of books in parallel, with a notably reduced per-message processing time. The implementation of our simulator - JAX-LOB - is based on design choices that aim to best exploit the powers of JAX without compromising on the realism of LOB-related mechanisms. We integrate JAX-LOB with other JAX packages, to provide an example of how one may address an optimal execution problem with reinforcement learning, and to share some preliminary results from end-to-end RL training on GPUs.

Alessandro Micheli, Silvia Sapora, Anthea Monod et al.

Many machine learning problems involve data supported on curved spaces such as spheres, rotation groups, hyperbolic spaces, and general Riemannian manifolds, where Euclidean geometry can distort distances, averages, and the resulting optimal transport (OT) problem. Existing manifold OT methods have pursued amortized out-of-sample maps, while entropic regularization has made discrete OT more scalable, but these advantages have remained largely disjoint. We propose Entropic Riemannian Neural Optimal Transport (Entropic RNOT), a unified framework that combines intrinsic entropic OT with amortized out-of-sample evaluation on Riemannian manifolds. Our method learns a single target-side Schrödinger potential through a neural pullback parameterization, recovers the induced Gibbs coupling, and uses the resulting conditional laws to construct intrinsic transport surrogates. These include barycentric projections on Cartan-Hadamard manifolds and heat-smoothed conditional surrogates on stochastically complete manifolds, the latter turning possibly atomic target laws into absolutely continuous ones. For fixed regularization $\varepsilon>0$, we prove that the proposed hypothesis class recovers the entropic optimal coupling in strong probabilistic metrics. As consequences, barycentric surrogates converge in $L^2$, while heat-smoothed surrogates are stable at fixed heat time and asymptotically unbiased as the heat time vanishes. The guarantees hold for compactly supported data on possibly noncompact manifolds. Empirically, our method matches or improves over Euclidean, tangent-space, and log-Euclidean baselines on benchmarks over $\mathbb{S}^2$, $\mathrm{SO}(3)$, $\mathrm{SPD}(3)$, $\mathrm{SE}(3)$, and $\mathbb{H}^2$, scales favorably relative to discrete manifold Sinkhorn, and in a protein-ligand docking application, refines poses on $\mathrm{SE}(3)$ without retraining or per-instance optimization.

Peer Nagy, Sascha Frey, Silvia Sapora et al.

Developing a generative model of realistic order flow in financial markets is a challenging open problem, with numerous applications for market participants. Addressing this, we propose the first end-to-end autoregressive generative model that generates tokenized limit order book (LOB) messages. These messages are interpreted by a Jax-LOB simulator, which updates the LOB state. To handle long sequences efficiently, the model employs simplified structured state-space layers to process sequences of order book states and tokenized messages. Using LOBSTER data of NASDAQ equity LOBs, we develop a custom tokenizer for message data, converting groups of successive digits to tokens, similar to tokenization in large language models. Out-of-sample results show promising performance in approximating the data distribution, as evidenced by low model perplexity. Furthermore, the mid-price returns calculated from the generated order flow exhibit a significant correlation with the data, indicating impressive conditional forecast performance. Due to the granularity of generated data, and the accuracy of the model, it offers new application areas for future work beyond forecasting, e.g. acting as a world model in high-frequency financial reinforcement learning applications. Overall, our results invite the use and extension of the model in the direction of autoregressive large financial models for the generation of high-frequency financial data and we commit to open-sourcing our code to facilitate future research.

Branton DeMoss, Silvia Sapora, Jakob Foerster et al.

We demonstrate the existence of a complexity phase transition in neural networks by studying the grokking phenomenon, where networks suddenly transition from memorization to generalization long after overfitting their training data. To characterize this phase transition, we introduce a theoretical framework for measuring complexity based on rate-distortion theory and Kolmogorov complexity, which can be understood as principled lossy compression for networks. We find that properly regularized networks exhibit a sharp phase transition: complexity rises during memorization, then falls as the network discovers a simpler underlying pattern that generalizes. In contrast, unregularized networks remain trapped in a high-complexity memorization phase. We establish an explicit connection between our complexity measure and generalization bounds, providing a theoretical foundation for the link between lossy compression and generalization. Our framework achieves compression ratios 30-40x better than naïve approaches, enabling precise tracking of complexity dynamics. Finally, we introduce a regularization method based on spectral entropy that encourages networks toward low-complexity representations by penalizing their intrinsic dimension.

Jonathan Cook, Silvia Sapora, Arash Ahmadian et al.

Training large language models (LLMs) on source code significantly enhances their general-purpose reasoning abilities, but the mechanisms underlying this generalisation are poorly understood. In this paper, we propose Programming by Backprop (PBB) as a potential driver of this effect - teaching a model to evaluate a program for inputs by training on its source code alone, without ever seeing I/O examples. To explore this idea, we finetune LLMs on two sets of programs representing simple maths problems and algorithms: one with source code and I/O examples (w/ IO), the other with source code only (w/o IO). We find evidence that LLMs have some ability to evaluate w/o IO programs for inputs in a range of experimental settings, and make several observations. Firstly, PBB works significantly better when programs are provided as code rather than semantically equivalent language descriptions. Secondly, LLMs can produce outputs for w/o IO programs directly, by implicitly evaluating the program within the forward pass, and more reliably when stepping through the program in-context via chain-of-thought. We further show that PBB leads to more robust evaluation of programs across inputs than training on I/O pairs drawn from a distribution that mirrors naturally occurring data. Our findings suggest a mechanism for enhanced reasoning through code training: it allows LLMs to internalise reusable algorithmic abstractions. Significant scope remains for future work to enable LLMs to more effectively learn from symbolic procedures, and progress in this direction opens other avenues like model alignment by training on formal constitutional principles.

Carlo Alfano, Silvia Sapora, Jakob Nicolaus Foerster et al.

Evaluating preference optimization (PO) algorithms on LLM alignment is a challenging task that presents prohibitive costs, noise, and several variables like model size and hyper-parameters. In this work, we show that it is possible to gain insights on the efficacy of PO algorithm on simpler benchmarks. We design a diagnostic suite of MuJoCo tasks and datasets, which we use to systematically evaluate PO algorithms, establishing a more controlled and cheaper benchmark. We then propose a novel family of PO algorithms based on mirror descent, which we call Mirror Preference Optimization (MPO). Through evolutionary strategies, we search this class to discover algorithms specialized to specific properties of preference datasets, such as mixed-quality or noisy data. We demonstrate that our discovered PO algorithms outperform all known algorithms in the targeted MuJoCo settings. Finally, based on the insights gained from our MuJoCo experiments, we design a PO algorithm that significantly outperform existing baselines in an LLM alignment task.

Carlo Alfano, Sebastian Towers, Silvia Sapora et al.

Policy Mirror Descent (PMD) is a popular framework in reinforcement learning, serving as a unifying perspective that encompasses numerous algorithms. These algorithms are derived through the selection of a mirror map and enjoy finite-time convergence guarantees. Despite its popularity, the exploration of PMD's full potential is limited, with the majority of research focusing on a particular mirror map -- namely, the negative entropy -- which gives rise to the renowned Natural Policy Gradient (NPG) method. It remains uncertain from existing theoretical studies whether the choice of mirror map significantly influences PMD's efficacy. In our work, we conduct empirical investigations to show that the conventional mirror map choice (NPG) often yields less-than-optimal outcomes across several standard benchmark environments. Using evolutionary strategies, we identify more efficient mirror maps that enhance the performance of PMD. We first focus on a tabular environment, i.e. Grid-World, where we relate existing theoretical bounds with the performance of PMD for a few standard mirror maps and the learned one. We then show that it is possible to learn a mirror map that outperforms the negative entropy in more complex environments, such as the MinAtar suite. Our results suggest that mirror maps generalize well across various environments, raising questions about how to best match a mirror map to an environment's structure and characteristics.

Silvia Sapora, Devon Hjelm, Alexander Toshev et al.

Inverse Reinforcement Learning aims to recover reward models from expert demonstrations, but traditional methods yield "black-box" models that are difficult to interpret and debug. In this work, we introduce GRACE (Generating Rewards As CodE), a method for using Large Language Models within an evolutionary search to reverse-engineer an interpretable, code-based reward function directly from expert trajectories. The resulting reward function is executable code that can be inspected and verified. We empirically validate GRACE on the BabyAI and AndroidWorld benchmarks, where it efficiently learns highly accurate rewards, even in complex, multi-task settings. Further, we demonstrate that the resulting reward leads to strong policies, compared to both competitive Imitation Learning and online RL approaches with ground-truth rewards. Finally, we show that GRACE is able to build complex reward APIs in multi-task setups.

Silvia Sapora, Gokul Swamy, Chris Lu et al.

Often times in imitation learning (IL), the environment we collect expert demonstrations in and the environment we want to deploy our learned policy in aren't exactly the same (e.g. demonstrations collected in simulation but deployment in the real world). Compared to policy-centric approaches to IL like behavioural cloning, reward-centric approaches like inverse reinforcement learning (IRL) often better replicate expert behaviour in new environments. This transfer is usually performed by optimising the recovered reward under the dynamics of the target environment. However, (a) we find that modern deep IL algorithms frequently recover rewards which induce policies far weaker than the expert, even in the same environment the demonstrations were collected in. Furthermore, (b) these rewards are often quite poorly shaped, necessitating extensive environment interaction to optimise effectively. We provide simple and scalable fixes to both of these concerns. For (a), we find that reward model ensembles combined with a slightly different training objective significantly improves re-training and transfer performance. For (b), we propose a novel evolution-strategies based method EvIL to optimise for a reward-shaping term that speeds up re-training in the target environment, closing a gap left open by the classical theory of IRL. On a suite of continuous control tasks, we are able to re-train policies in target (and source) environments more interaction-efficiently than prior work.

John Ahlgren, Maria Eugenia Berezin, Kinga Bojarczuk et al.

Software-intensive organizations rely on large numbers of software assets of different types, e.g., source-code files, tables in the data warehouse, and software configurations. Who is the most suitable owner of a given asset changes over time, e.g., due to reorganization and individual function changes. New forms of automation can help suggest more suitable owners for any given asset at a given point in time. By such efforts on ownership health, accountability of ownership is increased. The problem of finding the most suitable owners for an asset is essentially a program comprehension problem: how do we automatically determine who would be best placed to understand, maintain, evolve (and thereby assume ownership of) a given asset. This paper introduces the Facebook Ownesty system, which uses a combination of ultra large scale data mining and machine learning and has been deployed at Facebook as part of the company's ownership management approach. Ownesty processes many millions of software assets (e.g., source-code files) and it takes into account workflow and organizational aspects. The paper sets out open problems and challenges on ownership for the research community with advances expected from the fields of software engineering, programming languages, and machine learning.

John Ahlgren, Maria Eugenia Berezin, Kinga Bojarczuk et al.

We introduce the Web-Enabled Simulation (WES) research agenda, and describe FACEBOOK's WW system. We describe the application of WW to reliability, integrity and privacy at FACEBOOK , where it is used to simulate social media interactions on an infrastructure consisting of hundreds of millions of lines of code. The WES agenda draws on research from many areas of study, including Search Based Software Engineering, Machine Learning, Programming Languages, Multi Agent Systems, Graph Theory, Game AI, and AI Assisted Game Play. We conclude with a set of open problems and research challenges to motivate wider investigation.

Silvia Sapora, Bogdan Lazarescu, Christo Lolov

This document describes our approach to building an Offensive Language Classifier. More specifically, the OffensEval 2019 competition required us to build three classifiers with slightly different goals: - Offensive language identification: would classify a tweet as offensive or not. - Automatic categorization of offense types: would recognize if the target of the offense was an individual or not. - Offense target identification: would identify the target of the offense between an individual, group or other. In this report, we will discuss the different architectures, algorithms and pre-processing strategies we tried, together with a detailed description of the designs of our final classifiers and the reasons we choose them over others. We evaluated our classifiers on the official test set provided for the OffenseEval 2019 competition, obtaining a macro-averaged F1-score of 0.7189 for Task A, 0.6708 on Task B and 0.5442 on Task C.