Preventing Rank Collapse in Federated Low-Rank Adaptation with Client Heterogeneity

Federated low-rank adaptation (FedLoRA) has facilitated communication-efficient and privacy-preserving fine-tuning of foundation models for downstream tasks. In practical federated learning scenarios, client heterogeneity in system resources and data distributions motivates heterogeneous LoRA ranks across clients. We identify a previously overlooked phenomenon in heterogeneous FedLoRA, termed rank collapse, where the energy of the global update concentrates on the minimum shared rank, resulting in suboptimal performance and high sensitivity to rank configurations. Through theoretical analysis, we reveal the root cause of rank collapse: a mismatch between rank-agnostic aggregation weights and rank-dependent client contributions, which systematically suppresses higher-rank updates at a geometric rate over rounds. Motivated by this insight, we propose raFLoRA, a rank-partitioned aggregation method that decomposes local updates into rank partitions and then aggregates each partition weighted by its effective client contributions. Extensive experiments across classification and reasoning tasks show that raFLoRA prevents rank collapse, improves model performance, and preserves communication efficiency compared to state-of-the-art FedLoRA baselines.

Joint Continual Learning of Local Language Models and Cloud Offloading Decisions with Budget Constraints

Locally deployed Small Language Models (SLMs) must continually support diverse tasks under strict memory and computation constraints, making selective reliance on cloud Large Language Models (LLMs) unavoidable. Regulating cloud assistance during continual learning is challenging, as naive reward-based reinforcement learning often yields unstable offloading behavior and exacerbates catastrophic forgetting as task distributions shift. We propose DA-GRPO, a dual-advantage extension of Group Relative Policy Optimization that incorporates cloud-usage constraints directly into advantage computation, avoiding fixed reward shaping and external routing models. This design enables the local model to jointly learn task competence and collaboration behavior, allowing cloud requests to emerge naturally during post-training while respecting a prescribed assistance budget. Experiments on mathematical reasoning and code generation benchmarks show that DA-GRPO improves post-switch accuracy, substantially reduces forgetting, and maintains stable cloud usage compared to prior collaborative and routing-based approaches.

RCCDA: Adaptive Model Updates in the Presence of Concept Drift under a Constrained Resource Budget

Machine learning (ML) algorithms deployed in real-world environments are often faced with the challenge of adapting models to concept drift, where the task data distributions are shifting over time. The problem becomes even more difficult when model performance must be maintained under adherence to strict resource constraints. Existing solutions often depend on drift-detection methods that produce high computational overhead for resource-constrained environments, and fail to provide strict guarantees on resource usage or theoretical performance assurances. To address these shortcomings, we propose RCCDA: a dynamic model update policy that optimizes ML training dynamics while ensuring strict compliance to predefined resource constraints, utilizing only past loss information and a tunable drift threshold. In developing our policy, we analytically characterize the evolution of model loss under concept drift with arbitrary training update decisions. Integrating these results into a Lyapunov drift-plus-penalty framework produces a lightweight policy based on a measurable accumulated loss threshold that provably limits update frequency and cost. Experimental results on three domain generalization datasets demonstrate that our policy outperforms baseline methods in inference accuracy while adhering to strict resource constraints under several schedules of concept drift, making our solution uniquely suited for real-time ML deployments.

Dynamically Learned Test-Time Model Routing in Language Model Zoos with Service Level Guarantees

Open-weight LLM zoos provide access to numerous high-quality models, but selecting the appropriate model for specific tasks remains challenging and requires technical expertise. Most users simply want factually correct, safe, and satisfying responses without concerning themselves with model technicalities, while inference service providers prioritize minimizing operating costs. These competing interests are typically mediated through service level agreements (SLAs) that guarantee minimum service quality. We introduce MESS+, a stochastic optimization algorithm for cost-optimal LLM request routing while providing rigorous SLA compliance guarantees. MESS+ learns request satisfaction probabilities of LLMs in real-time as users interact with the system, based on which model selection decisions are made by solving a per-request optimization problem. Our algorithm includes a novel combination of virtual queues and request satisfaction prediction, along with a theoretical analysis of cost optimality and constraint satisfaction. Across a wide range of state-of-the-art LLM benchmarks, MESS+ achieves an average of 2x cost savings compared to existing LLM routing techniques.

Memory-Efficient Orthogonal Fine-Tuning with Principal Subspace Adaptation

Driven by the relentless growth in model parameters, which renders full fine-tuning prohibitively expensive for large-scale deployment, parameter-efficient fine-tuning (PEFT) has emerged as a crucial approach for rapidly adapting large models to a wide range of downstream tasks. Among the PEFT family, orthogonal fine-tuning and its variants have demonstrated remarkable performance by preserving hyperspherical energy, which encodes pairwise angular similarity between neurons. However, these methods are inherently memory-inefficient due to the need to store intermediate activations from multiple full-dimensional sparse matrices. To address this limitation, we propose Memory-efficient Orthogonal Fine-Tuning (MOFT) with principal subspace adaptation. Specifically, we first establish a theoretical condition under which orthogonal transformations within a low-rank subspace preserve hyperspherical energy. Based on this insight, we constrain orthogonal fine-tuning to the principal subspace defined by the top-r components obtained through singular value decomposition and impose an additional constraint on the projection matrix to satisfy the preservation condition. To enhance MOFT's flexibility across tasks, we relax strict orthogonality by introducing two learnable scaling vectors. Extensive experiments on 37 diverse tasks and four models across NLP and CV demonstrate that MOFT consistently outperforms key baselines while significantly reducing the memory footprint of orthogonal fine-tuning.

IPBench: Benchmarking the Knowledge of Large Language Models in Intellectual Property

Intellectual Property (IP) is a unique domain that integrates technical and legal knowledge, making it inherently complex and knowledge-intensive. As large language models (LLMs) continue to advance, they show great potential for processing IP tasks, enabling more efficient analysis, understanding, and generation of IP-related content. However, existing datasets and benchmarks either focus narrowly on patents or cover limited aspects of the IP field, lacking alignment with real-world scenarios. To bridge this gap, we introduce the first comprehensive IP task taxonomy and a large, diverse bilingual benchmark, IPBench, covering 8 IP mechanisms and 20 tasks. This benchmark is designed to evaluate LLMs in real-world intellectual property applications, encompassing both understanding and generation. We benchmark 16 LLMs, ranging from general-purpose to domain-specific models, and find that even the best-performing model achieves only 75.8% accuracy, revealing substantial room for improvement. Notably, open-source IP and law-oriented models lag behind closed-source general-purpose models. We publicly release all data and code of IPBench and will continue to update it with additional IP-related tasks to better reflect real-world challenges in the intellectual property domain.

GneissWeb: Preparing High Quality Data for LLMs at Scale

Data quantity and quality play a vital role in determining the performance of Large Language Models (LLMs). High-quality data, in particular, can significantly boost the LLM's ability to generalize on a wide range of downstream tasks. Large pre-training datasets for leading LLMs remain inaccessible to the public, whereas many open datasets are small in size (less than 5 trillion tokens), limiting their suitability for training large models. In this paper, we introduce GneissWeb, a large dataset yielding around 10 trillion tokens that caters to the data quality and quantity requirements of training LLMs. Our GneissWeb recipe that produced the dataset consists of sharded exact sub-string deduplication and a judiciously constructed ensemble of quality filters. GneissWeb achieves a favorable trade-off between data quality and quantity, producing models that outperform models trained on state-of-the-art open large datasets (5+ trillion tokens). We show that models trained using GneissWeb dataset outperform those trained on FineWeb-V1.1.0 by 2.73 percentage points in terms of average score computed on a set of 11 commonly used benchmarks (both zero-shot and few-shot) for pre-training dataset evaluation. When the evaluation set is extended to 20 benchmarks (both zero-shot and few-shot), models trained using GneissWeb still achieve a 1.75 percentage points advantage over those trained on FineWeb-V1.1.0.

Dynamic Loss-Based Sample Reweighting for Improved Large Language Model Pretraining

Pretraining large language models (LLMs) on vast and heterogeneous datasets is crucial for achieving state-of-the-art performance across diverse downstream tasks. However, current training paradigms treat all samples equally, overlooking the importance or relevance of individual samples throughout the training process. Existing reweighting strategies, which primarily focus on group-level data importance, fail to leverage fine-grained instance-level information and do not adapt dynamically to individual sample importance as training progresses. In this paper, we introduce novel algorithms for dynamic, instance-level data reweighting aimed at improving both the efficiency and effectiveness of LLM pretraining. Our methods adjust the weight of each training sample based on its loss value in an online fashion, allowing the model to dynamically focus on more informative or important samples at the current training stage. In particular, our framework allows us to systematically devise reweighting strategies deprioritizing redundant or uninformative data, which we find tend to work best. Furthermore, we develop a new theoretical framework for analyzing the impact of loss-based reweighting on the convergence of gradient-based optimization, providing the first formal characterization of how these strategies affect convergence bounds. We empirically validate our approach across a spectrum of tasks, from pretraining 7B and 1.4B parameter LLMs to smaller-scale language models and linear regression problems, demonstrating that our loss-based reweighting approach can lead to faster convergence and significantly improved performance.

Parameter Tracking in Federated Learning with Adaptive Optimization

In Federated Learning (FL), model training performance is strongly impacted by data heterogeneity across clients. Gradient Tracking (GT) has recently emerged as a solution which mitigates this issue by introducing correction terms to local model updates. To date, GT has only been considered under Stochastic Gradient Descent (SGD)-based model training, while modern FL frameworks increasingly employ adaptive optimizers for improved convergence. In this work, we generalize the GT framework to a more flexible Parameter Tracking (PT) paradigm and propose two novel adaptive optimization algorithms, {\tt FAdamET} and {\tt FAdamGT}, that integrate PT into Adam-based FL. We provide a rigorous convergence analysis of these algorithms under non-convex settings. Our experimental results demonstrate that both proposed algorithms consistently outperform existing methods when evaluating total communication cost and total computation cost across varying levels of data heterogeneity, showing the effectiveness of correcting first-order information in federated adaptive optimization.

Adaptive Rank Allocation for Federated Parameter-Efficient Fine-Tuning of Language Models

Pre-trained Language Models (PLMs) have demonstrated their superiority and versatility in modern Natural Language Processing (NLP), effectively adapting to various downstream tasks through further fine-tuning. Federated Parameter-Efficient Fine-Tuning (FedPEFT) has emerged as a promising solution to address privacy and efficiency challenges in distributed training for PLMs on mobile devices. However, our measurements reveal two key limitations of FedPEFT: heterogeneous data leads to significant performance degradation, and a fixed parameter configuration results in communication inefficiency. To overcome these limitations, we propose FedARA, a novel Federated Adaptive Rank Allocation for parameter-efficient fine-tuning of language models. Specifically, FedARA employs truncated singular value decomposition (SVD) adaptation to enhance flexibility and expressiveness, significantly mitigating the adverse effects of data heterogeneity. Subsequently, it utilizes dynamic rank allocation to progressively identify critical ranks, effectively improving communication efficiency. Lastly, it leverages rank-based module pruning to remove inactive modules, steadily reducing local training time and peak memory usage in each round. Extensive experiments show that FedARA consistently outperforms weak baselines by an average of 8.49\% and strong baselines by 6.95\% across various datasets under data heterogeneity while significantly improving communication efficiency by 2.40\(\times\). Moreover, experiments on AGX Orin, Orin Nano and Raspberry Pi 5 devices demonstrate substantial decreases in total training time and energy consumption by up to 48.90\% and 46.95\%, respectively.