Amin Banayeeanzade, Ala N. Tak, Fatemeh Bahrani, Anahita Bolourani, Leonardo Blas, Emilio Ferrara, Jonathan Gratch, Sai Praneeth Karimireddy. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

The problem of model privacy against an eavesdropper (Eve) in a distributed learning environment is investigated. The solution is found via evaluating the Fisher Information Matrix (FIM) for the model learning problem for Eve. Through a model shift design process, the eavesdropper’s FIM can be driven to singularity, yielding a provably hard estimation problem for Eve. Both a one-shot and multi-shot solution are designed. These two approaches require the sharing of a modest amount of information with the central server learning the global model. The multi-shot solution has time-varying shifts that prevent Eve from using the temporal correlation of the gradients to learn the shifts. We design a convergence test for Eve to determine if model updates have been tampered with. However, our shift strategies pass the test and thus the shifts are not detectable. The single-shot and multi-shot methods are compared against a noise injection scheme and shown to offer superior performance.

Yavuz Faruk Bakman, Duygu Nur Yaldiz, Sungmin Kang, Tuo Zhang, Baturalp Buyukates, Salman Avestimehr, Sai Praneeth Karimireddy. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

Reinforcement learning from human feedback (RLHF) and, at its core, reward modeling have become a crucial part of training powerful large language models (LLMs).One commonly overlooked factor in training high-quality reward models (RMs) is the effect of the base model, which is becoming more challenging to choose given the rapidly growing pool of LLMs.In this work, we present a systematic analysis of the effect of base model selection on reward modeling performance.Our results show that the performance can be improved by up to 14% compared to the most common (i.e., default) choice.Moreover, we showcase the strong statistical relation between some existing benchmarks and downstream performances.We also demonstrate that the results from a small set of benchmarks could be combined to boost the model selection (+18% on average in the top 5-10).Lastly, we illustrate the impact of different post-training steps on the final performance and explore using estimated data distributions to reduce performance prediction error.

In Federated Learning, it is crucial to handle low-quality, corrupted, or malicious data. However, traditional data valuation methods are not suitable due to privacy concerns. To address this, we propose a simple yet effective approach that utilizes a new influence approximation called "lazy influence" to filter and score data while preserving privacy. To do this, each participant uses their own data to estimate the influence of another participant’s batch and sends a differentially private obfuscated score to the central coordinator. Our method has been shown to successfully filter out biased and corrupted data in various simulated and real-world settings, achieving a recall rate of over > 90% (sometimes up to 100%) while maintaining strong differential privacy guarantees with ε ≤ 1.

Collaboration between different data centers is often challenged by heterogeneity across sites. To account for the heterogeneity, the state-of-the-art method is to re-weight the covariate distributions in each site to match the distribution of the target population. Nevertheless, this method still relies on the concept of traditional meta-analysis after adjusting for the distribution shift. This work proposes a collaborative inverse propensity score weighting estimator for causal inference with heterogeneous data. Instead of adjusting the distribution shift separately, we use weighted propensity score models to collaboratively adjust for the distribution shift. Our method shows significant improvements over the methods based on meta-analysis when heterogeneity increases. By incorporating outcome regression models, we prove the asymptotic normality when the covariates have dimension $d<8$. Our methods preserve privacy at individual sites by implementing federated learning protocols.

The acquisition of training data is crucial for machine learning applications. Data markets can increase the supply of data, particularly in data-scarce domains such as healthcare, by incentivizing potential data providers to join the market. A major challenge for a data buyer in such a market is choosing the most valuable data points from a data seller. Unlike prior work in data valuation, which assumes centralized data access, we propose a federated approach to the data acquisition problem that is inspired by linear experimental design. Our proposed data acquisition method achieves lower prediction error without requiring labeled validation data and can be optimized in a fast and federated procedure. The key insight of our work is that a method that directly estimates the benefit of acquiring data for test set prediction is particularly compatible with a decentralized market setting.

Conformal prediction is emerging as a popular paradigm for providing rigorous uncertainty quantification in machine learning since it can be easily applied as a post-processing step to already trained models. In this paper, we extend conformal prediction to the federated learning setting. The main challenge we face is data heterogeneity across the clients --- this violates the fundamental tenet of *exchangeability* required for conformal prediction. We propose a weaker notion of *partial exchangeability*, better suited to the FL setting, and use it to develop the Federated Conformal Prediction (FCP) framework. We show FCP enjoys rigorous theoretical guarantees and excellent empirical performance on several computer vision and medical imaging datasets. Our results demonstrate a practical approach to incorporating meaningful uncertainty quantification in distributed and heterogeneous environments. We provide code used in our experiments https://github.com/clu5/federated-conformal.

Federated Learning (FL) is a novel approach enabling several clients holding sensitive data to collaboratively train machine learning models, without centralizing data. The cross-silo FL setting corresponds to the case of few ($2$--$50$) reliable clients, each holding medium to large datasets, and is typically found in applications such as healthcare, finance, or industry. While previous works have proposed representative datasets for cross-device FL, few realistic healthcare cross-silo FL datasets exist, thereby slowing algorithmic research in this critical application. In this work, we propose a novel cross-silo dataset suite focused on healthcare, FLamby (Federated Learning AMple Benchmark of Your cross-silo strategies), to bridge the gap between theory and practice of cross-silo FL.FLamby encompasses 7 healthcare datasets with natural splits, covering multiple tasks, modalities, and data volumes, each accompanied with baseline training code. As an illustration, we additionally benchmark standard FL algorithms on all datasets.Our flexible and modular suite allows researchers to easily download datasets, reproduce results and re-use the different components for their research. FLamby is available at~\url{www.github.com/owkin/flamby}.

State-of-the-art federated learning methods can perform far worse than their centralized counterparts when clients have dissimilar data distributions. For neural networks, even when centralized SGD easily finds a solution that is simultaneously performant for all clients, current federated optimization methods fail to converge to a comparable solution. We show that this performance disparity can largely be attributed to optimization challenges presented by nonconvexity. Specifically, we find that the early layers of the network do learn useful features, but the final layers fail to make use of them. That is, federated optimization applied to this non-convex problem distorts the learning of the final layers. Leveraging this observation, we propose a Train-Convexify-Train (TCT) procedure to sidestep this issue: first, learn features using off-the-shelf methods (e.g., FedAvg); then, optimize a convexified problem obtained from the network's empirical neural tangent kernel approximation. Our technique yields accuracy improvements of up to $+36\%$ on FMNIST and $+37\%$ on CIFAR10 when clients have dissimilar data.

Decentralized training of deep learning models is a key element for enabling data privacy and on-device learning over networks. In realistic learning scenarios, the presence of heterogeneity across different clients’ local datasets poses an optimization challenge and may severely deteriorate the generalization performance. In this paper, we investigate and identify the limitation of several decentralized optimization algorithms for different degrees of data heterogeneity. We propose a novel momentum-based method to mitigate this decentralized training difficulty. We show in extensive empirical experiments on various CV/NLP datasets (CIFAR-10, ImageNet, and AG News) and several network topologies (Ring and Social Network) that our method is much more robust to the heterogeneity of clients’ data than other existing methods, by a significant improvement in test performance (1%-20%).

Byzantine robustness has received significant attention recently given its importance for distributed and federated learning. In spite of this, we identify severe flaws in existing algorithms even when the data across the participants is identically distributed. First, we show realistic examples where current state of the art robust aggregation rules fail to converge even in the absence of any Byzantine attackers. Secondly, we prove that even if the aggregation rules may succeed in limiting the influence of the attackers in a single round, the attackers can couple their attacks across time eventually leading to divergence. To address these issues, we present two surprisingly simple strategies: a new robust iterative clipping procedure, and incorporating worker momentum to overcome time-coupled attacks. This is the first provably robust method for the standard stochastic optimization setting.

Federated learning (FL) is a challenging setting for optimization due to the heterogeneity of the data across different clients which gives rise to the client drift phenomenon. In fact, obtaining an algorithm for FL which is uniformly better than simple centralized training has been a major open problem thus far. In this work, we propose a general algorithmic framework, Mime, which i) mitigates client drift and ii) adapts arbitrary centralized optimization algorithms such as momentum and Adam to the cross-device federated learning setting. Mime uses a combination of control-variates and server-level statistics (e.g. momentum) at every client-update step to ensure that each local update mimics that of the centralized method run on iid data. We prove a reduction result showing that Mime can translate the convergence of a generic algorithm in the centralized setting into convergence in the federated setting. Further, we show that when combined with momentum based variance reduction, Mime is provably faster than any centralized method--the first such result. We also perform a thorough experimental exploration of Mime's performance on real world datasets.

In decentralized machine learning, workers compute model updates on their local data.Because the workers only communicate with few neighbors without central coordination, these updates propagate progressively over the network.This paradigm enables distributed training on networks without all-to-all connectivity, helping to protect data privacy as well as to reduce the communication cost of distributed training in data centers.A key challenge, primarily in decentralized deep learning, remains the handling of differences between the workers' local data distributions.To tackle this challenge, we introduce the RelaySum mechanism for information propagation in decentralized learning.RelaySum uses spanning trees to distribute information exactly uniformly across all workers with finite delays depending on the distance between nodes.In contrast, the typical gossip averaging mechanism only distributes data uniformly asymptotically while using the same communication volume per step as RelaySum.We prove that RelaySGD, based on this mechanism, is independent of data heterogeneity and scales to many workers, enabling highly accurate decentralized deep learning on heterogeneous data.

Gradient Boosting Machine (GBM) introduced by \cite{friedman2001greedy} is a widely popular ensembling technique and is routinely used in competitions such as Kaggle and the KDDCup \citep{chen2016xgboost}. In this work, we propose an Accelerated Gradient Boosting Machine (AGBM) by incorporating Nesterov’s acceleration techniques into the design of GBM. The difficulty in accelerating GBM lies in the fact that weak (inexact) learners are commonly used, and therefore, with naive application, the errors can accumulate in the momentum term. To overcome it, we design a “corrected pseudo residual” that serves as a new target for fitting a weak learner, in order to perform the z-update. Thus, we are able to derive novel computational guarantees for AGBM. This is the first GBM type of algorithm with a theoretically-justified accelerated convergence rate.

Federated learning is a key scenario in modern large-scale machine learning where the data remains distributed over a large number of clients and the task is to learn a centralized model without transmitting the client data. The standard optimization algorithm used in this setting is Federated Averaging (FedAvg) due to its low communication cost. We obtain a tight characterization of the convergence of FedAvg and prove that heterogeneity (non-iid-ness) in the client’s data results in a ‘drift’ in the local updates resulting in poor performance. As a solution, we propose a new algorithm (SCAFFOLD) which uses control variates (variance reduction) to correct for the ‘client drift’. We prove that SCAFFOLD requires significantly fewer communication rounds and is not affected by data heterogeneity or client sampling. Further, we show that (for quadratics) SCAFFOLD can take advantage of similarity in the client’s data yielding even faster convergence. The latter is the first result to quantify the usefulness of local-steps in distributed optimization.

Lossy gradient compression has become a practical tool to overcome the communication bottleneck in centrally coordinated distributed training of machine learning models. However, algorithms for decentralized training with compressed communication over arbitrary connected networks have been more complicated, requiring additional memory and hyperparameters. We introduce a simple algorithm that directly compresses the model differences between neighboring workers using low-rank linear compressors. We prove that our method does not require any additional hyperparameters, converges faster than prior methods, and is asymptotically independent of both the network and the compression. Inspired the PowerSGD algorithm for centralized deep learning, we execute power iteration steps on model differences to maximize the information transferred per bit. Out of the box, these compressors perform on par with state-of-the-art tuned compression algorithms in a series of deep learning benchmarks.

While stochastic gradient descent (SGD) is still the de facto algorithm in deep learning, adaptive methods like Clipped SGD/Adam have been observed to outperform SGD across important tasks, such as attention models. The settings under which SGD performs poorly in comparison to adaptive methods are not well understood yet. In this paper, we provide empirical and theoretical evidence that a heavy-tailed distribution of the noise in stochastic gradients is one cause of SGD's poor performance. We provide the first tight upper and lower convergence bounds for adaptive gradient methods under heavy-tailed noise. Further, we demonstrate how gradient clipping plays a key role in addressing heavy-tailed gradient noise. Subsequently, we show how clipping can be applied in practice by developing an adaptive coordinate-wise clipping algorithm (ACClip) and demonstrate its superior performance on BERT pretraining and finetuning tasks.

Coordinate descent with random coordinate selection is the current state of the art for many large scale optimization problems. However, greedy selection of the steepest coordinate on smooth problems can yield convergence rates independent of the dimension $n$, requiring $n$ times fewer iterations. In this paper, we consider greedy updates that are based on subgradients for a class of non-smooth composite problems, including $L1$-regularized problems, SVMs and related applications. For these problems we provide (i) the first linear rates of convergence independent of $n$, and show that our greedy update rule provides speedups similar to those obtained in the smooth case. This was previously conjectured to be true for a stronger greedy coordinate selection strategy. Furthermore, we show that (ii) our new selection rule can be mapped to instances of maximum inner product search, allowing to leverage standard nearest neighbor algorithms to speed up the implementation. We demonstrate the validity of the approach through extensive numerical experiments.

Sign-based algorithms (e.g. signSGD) have been proposed as a biased gradient compression technique to alleviate the communication bottleneck in training large neural networks across multiple workers. We show simple convex counter-examples where signSGD does not converge to the optimum. Further, even when it does converge, signSGD may generalize poorly when compared with SGD. These issues arise because of the biased nature of the sign compression operator. We then show that using error-feedback, i.e. incorporating the error made by the compression operator into the next step, overcomes these issues. We prove that our algorithm (EF-SGD) with arbitrary compression operator achieves the same rate of convergence as SGD without any additional assumptions. Thus EF-SGD achieves gradient compression for free. Our experiments thoroughly substantiate the theory.

We study gradient compression methods to alleviate the communication bottleneck in data-parallel distributed optimization. Despite the significant attention received, current compression schemes either do not scale well, or fail to achieve the target test accuracy. We propose a low-rank gradient compressor that can i) compress gradients rapidly, ii) efficiently aggregate the compressed gradients using all-reduce, and iii) achieve test performance on par with SGD. The proposed algorithm is the only method evaluated that achieves consistent wall-clock speedups when benchmarked against regular SGD with an optimized communication backend. We demonstrate reduced training times for convolutional networks as well as LSTMs on common datasets.

First-order optimization methods comprise two important primitives: i) the computation of gradient information and ii) the computation of the update that leads to the next iterate. In practice there is often a wide mismatch between the time required for the two steps, leading to underutilization of resources. In this work, we propose a new framework, Approx Composite Minimization (ACM) that uses approximate update steps to ensure balance between the two operations. The accuracy is adaptively chosen in an online fashion to take advantage of changing conditions. Our unified analysis for approximate composite minimization generalizes and extends previous work to new settings. Numerical experiments on Lasso regression and SVMs demonstrate the effectiveness of the novel scheme.

Two popular examples of first-order optimization methods over linear spaces are coordinate descent and matching pursuit algorithms, with their randomized variants. While the former targets the optimization by moving along coordinates, the latter considers a generalized notion of directions. Exploiting the connection between the two algorithms, we present a unified analysis of both, providing affine invariant sublinear $O(1/t)$ rates on smooth objectives and linear convergence on strongly convex objectives. As a byproduct of our affine invariant analysis of matching pursuit, our rates for steepest coordinate descent are the tightest known. Furthermore, we show the first accelerated convergence rate $O(1/t^2)$ for matching pursuit and steepest coordinate descent on convex objectives.