Gradient-Free Rank Allocator

Model merging aims to combine multiple fine-tuned models derived from a common pretrained model into a single multi-task model without additional joint training. Recent spectral merging methods improve over simple weight averaging by exploiting low-rank structures of task-specific updates, but they commonly assign the same rank capacity to every task. This uniform allocation ignores that task vectors can have heterogeneous spectral complexity, causing the shared merging space to be used suboptimally. In this paper, we propose Spectral Energy-proportional Rank Allocation (SERA), a simple task-adaptive strategy that allocates ranks according to the singular-value energy structure of each task vector. By assigning richer spectral capacity to complex or isolated tasks and fewer directions to compact tasks, SERA extends SVD-based model merging from uniform-capacity merging to task-dependent capacity allocation. Experiments under standard vision model merging protocols show that SERA improves multi-task merging performance while preserving the same total rank budget as existing spectral merging methods. Further analysis demonstrates that task-level spectral concentration is closely related to the per-task effect of adaptive rank allocation, providing insight into when and why SERA is effective.

Low-rank gradient projection reduces the optimizer-state memory cost of large language model (LLM) pretraining, but the steps and wall-clock time needed to reach a target quality remain a meaningful axis for improvement. We attribute this to two design choices in existing methods: the projection-rank budget is allocated uniformly across Transformer modules with heterogeneous projection sensitivity, and projecting a raw gradient attenuates its magnitude and direction jointly. We propose MoARa, which combines a static profiling-based module-aware projection-rank allocation with a block-wise magnitude-direction decomposition; the default block size is set in the neighborhood of the attention head dimension. Across five Transformer architectures spanning Llama, Qwen, and DeepSeek at 300M to 7B scales, GaLore with MoARa reaches standard GaLore's final perplexity in 37% fewer steps and 34% less wall-clock time on Llama 2 7B, with only 0.2% peak reserved memory overhead under standard graph compilation. Across the six low-rank pretraining methods we evaluate, module-aware rank allocation alone delivers directionally consistent step reductions on all six. On compatible hosts, the two-component design reaches up to 41.7% step reduction and 37.1% wall-clock reduction.

Offline multi-agent payoff models are estimated under a logging distribution but used on distributions induced by learned solutions and unilateral deviations. Standard held-out loss can therefore favor an interaction class that predicts logged play well while distorting strategic incentives. We introduce Selective Interaction-Rank Validation (SIRV) for finite games with known logging distributions. A training split fits nested payoff models and constructs a common union of all candidate deployment and unilateral-replacement distributions; an independent calibration split evaluates every candidate on this same union. SIRV returns the smallest rank whose simultaneous upper worst-target risk is within tolerance of the best upper score, and abstains when a declared target is unsupported or too imprecisely estimated. A common coverage event yields a finite-candidate target-risk bound and a candidate-specific coarse correlated equilibrium (CCE) gap certificate. We also isolate an exact two-point off-support non-identifiability result. In a controlled factorial study with 2,048 independent games per family, empirical-Bernstein bounds reduce the median CCE-gap certificate by 42.5% relative to Hoeffding bounds on common returns, with a 1.36-point reduction in supported return. Under paired rank misspecification and in a separately generated congestion family, the SIRV-EB fallback rule lowers mean true candidate-selection CCE regret relative to ID-Mean, while retaining game-level losses. Across 384 games at $N=3,5,8$, ID-Mean-relative mean CCE-regret effects stay positive while certified return falls sharply under weak coverage. These results separate certifiable model selection from universal strategic improvement.

Low-Rank Adaptation (LoRA) is a prominent fine-tuning method for large models, achieving competitive performance with reduced memory overhead. However, a persistent performance gap remains between LoRA and full fine-tuning. Recent studies have sought to narrow this gap by employing one-step gradient approximations of pretrained weights to align LoRA updates with the principal directions or intrinsic dimensionalities of full fine-tuning updates. Nevertheless, these approaches fail to capture the full dynamics of the gradients. In this paper, we propose LoRA-GA$^2$, an effective fine-tuning algorithm that fully leverages multi-step gradient information. Specifically, we introduce a lightweight probe for multi-step gradients of pretrained weights that incurs no additional GPU memory cost and only marginal time overhead. We further employ a spectrum-aware, importance-based rank allocation and optimal initialization derived from multi-step gradients. Extensive experimental results demonstrate that LoRA-GA$^2$ consistently outperforms existing LoRA variants while preserving the efficiency advantages of vanilla LoRA. For instance, LoRA-GA$^2$ surpasses the leading baseline by an average of 0.66 points on the GLUE benchmark, and outperforms the strongest baseline by 1.03 points on GSM8K and 0.87 points on HumanEval, respectively.

As the inference phase of Large Language Models (LLMs) requires handling long context windows, the Key-Value (KV) cache initially appears to address this challenge but eventually becomes a significant bottleneck as the context window continues to grow. Low-rank compression has recently been studied as an effective approach to reduce KV cache memory while maintaining model performance. However, only a few existing methods treat the Key and Value caches differently, despite their distinct roles. Moreover, these methods typically employ fixed attention-head grouping, which may not fully exploit the structural similarity among attention heads. In this paper, we propose an improved low-rank KV cache compression framework. For the Key cache, we dynamically group attention heads based on Centered Kernel Alignment (CKA) similarity and allocate the rank budget adaptively under a parameter budget. For the Value cache, we adopt the same approach as ReCalKV, refining the low-rank decomposition through offline calibration to improve reconstruction quality. Experimental results on three instruction-tuned LLMs show that our method reduces the number of Key cache parameters while maintaining competitive accuracy. We further observe that the proposed strategy is particularly effective for Multi-Head Attention (MHA) models, whereas it should be applied more conservatively to Grouped-Query Attention (GQA) models, especially in long-context settings.

Low-Rank Adaptation (LoRA) is a widely used approach to parameter-efficient fine-tuning (PEFT) of LLMs whose effectiveness depends on rank allocation. Existing adaptive LoRA methods derive ranks from local gradient, activation, or matrix statistics collected before or during fine-tuning; training-time variants add overhead, and local signals reveal little about each module's structural role in information propagation, giving weak global grounding for scarce-capacity allocation. We propose IFCLoRA, a topology-aware method for pre-fine-tuning rank allocation and adapter initialization. Using a small calibration set, IFCLoRA performs intervention tracing on the frozen model and constructs a sparse task-conditioned interaction graph over LoRA target modules. From this graph it extracts a global information-flow topology prior and fuses it with each node's local gradient sensitivity to form a topology-dominant Information-Flow Centrality (IFC) score, measuring participation in task-conditioned multi-hop propagation. The IFC scores then serve as module-level routing signals for one-shot discrete rank allocation under a rank-budget constraint. Reusing response vectors from tracing, IFCLoRA constructs a function-preserving flow-response subspace initialization, giving adapters task-relevant output subspaces. Across all settings, IFCLoRA achieves higher mean scores than standard LoRA with comparable fine-tuning time and peak memory; it requires a one-time offline calibration stage. On GSM8K, IFCLoRA attains the highest mean accuracy among compared PEFT methods on both base models, exceeding standard LoRA by 4.75 percentage points on LLaMA-3.1-8B. Resulting rank allocations are non-uniform and vary across tasks and base models, suggesting that task-conditioned global information-flow topology can serve as a useful structural prior for rank allocation in low-budget PEFT.

Low-rank adaptation (LoRA) has become a widely used parameter-efficient fine-tuning method for large language models. Since different modules and layers may contribute unequally to downstream adaptation, allocating rank resources under a fixed parameter budget is an important problem for balancing efficiency, expressiveness, and generalization. Existing adaptive rank methods address this problem mainly through carefully designed importance scores constructed from gradient-derived sensitivity and uncertainty measures, without an explicit statistical interpretation. In this paper, we formulate LoRA rank allocation as a statistical hypothesis testing problem and propose StatLoRA, a statistical inference-based rank allocation method. StatLoRA associates each LoRA component with a test statistic and uses estimated p-values to determine which components should be retained or pruned under a prescribed rank budget. The proposed testing procedure is supported by our central limit theory for stochastic optimizer trajectories. In particular, we establish asymptotic normality for a broad class of commonly used optimizers in deep learning, including AdamW, and derive the corresponding asymptotic distributions for the proposed component scores used in hypothesis testing. We evaluate StatLoRA on LoRA fine-tuning of DeBERTaV3-base, BART-Large, and Qwen2.5-7B across natural language understanding, natural language generation, and question answering tasks. Experiments show that StatLoRA achieves comparable or better performance than vanilla LoRA, AdaLoRA, and IGU-LoRA under matched rank budgets. Sensitivity analyses and empirical diagnostics further support the stability of the proposed hypothesis-testing-based allocation rule and provide empirical evidence for the asymptotic theory of component scores.

Low-Rank Adaptation (LoRA) has become a cornerstone of parameter-efficient fine-tuning (PEFT); however, the conventional practice of uniform rank assignment ignores the functional heterogeneity of neural layers. Existing rank allocation methods typically struggle with a trade-off between computational intensity and heuristic simplicity: training-based methods suffer from prohibitive overhead, while pre-allocation methods fail to capture the dynamic task-specific representation manifold. In this paper, we propose RSLoRA (Representational Sensitivity LoRA), a training-free and gradient-free rank allocator driven by activation-space geometry. We identify a "sensitivity regime shift" across layers, observing that static weight analysis and local gradients are insufficient to reflect how updates reshape a model's internal representations. To address this, RSLoRA introduces a virtual representational probing mechanism. By simulating adaptation through structured low-rank noise and measuring the resulting manifold displacement by using Effective Rank and Frechet Distance, we identify high-sensitivity modules that require higher rank capacity. Our framework effectively bridges the gap between expert-crafted heuristics and actual representational impact. Extensive evaluations demonstrate that RSLoRA consistently outperforms state-of-the-art allocators (e.g., AdaLoRA, GoRA) across mainstream benchmarks. By eliminating the need for iterative training-time adjustments and backward gradients, RSLoRA provides a highly efficient, robust, and representation-aware solution for large-scale model adaptation.

Low-rank decomposition is a promising compression paradigm for large language models (LLMs), yet its effectiveness hinges on rank budget allocation across weight matrices: uniform or hand-crafted rules ignore module-wise importance, while learning-based allocation incurs substantial training overhead. We formulate rank allocation as a global sorting-and-truncation pipeline that scores every singular component by combining local singular energy with global functional importance, estimated via layer-wise input--output cosine similarity on a tiny calibration set. We show, both geometrically and empirically, that high input--output cosine similarity implies low effective rank. We further propose rank-preserving fine-tuning (RPFT), which adapts only a small subset of retained singular components so that the allocated rank stays bounded without re-decomposition. Experimental results show that UniRank cuts zero-shot perplexity by up to 50%, improves average reasoning accuracy by 3.0% over LoRAP at 25% sparsity, and boosts four SVD-based decomposition methods as a plug-and-play module.

We propose annealed entropic allocation, an adaptive sampling policy based on an annealed, weighted soft-min formulation of static budget allocation. We replace the maximin large-deviation rate objective with a weighted log-sum-exp surrogate that blends challenger-specific pairwise scores through soft-min weights, avoiding hard switching when several challengers are nearly active. To capture tail behavior beyond the leading exponent, the surrogate incorporates saddlepoint prefactors from refined pairwise tail asymptotics. Because these corrections are subexponential, decreasing the annealing temperature with the budget preserves the same first-order target allocation. For the static problem, we prove uniform convergence to the hard minimum, concentration of soft-min weights on active challengers, and continuity of the induced target-allocation map under fixed weights. Experiments show that the proposed methods are consistently competitive: the no-saddlepoint ablation performs best in symmetric Gaussian and exponential slippage settings, while saddlepoint weighting can help in heterogeneous or asymmetric cases.

The massive scaling of Large Language Models (LLMs) has made pretraining increasingly cost-prohibitive. While low-rank representation and orthonormal weight matrices could in principle reduce parameter counts and computational overhead, most existing methods rely on static rank selection and do not enforce weight orthonormality due to high computational cost. This paper introduces TSVD, a framework that maintains low rank and strict orthonormality throughout the training process. It utilizes a spectral energy-based heuristic for adaptive rank selection, and a caching mechanisms to maintain orthonormality. Theoretical analysis justifies the advantage of the approach in pretraining dynamics and experiments across various model scales demonstrate that it is effective empirically. TSVD matches or exceeds the performance of full-parameter baselines while significantly reducing compute requirements. The approach thus offers a well-founded, practical, and scalable path toward efficient high-performance LLM pretraining.

Low-rank adaptation (LoRA) assigns a uniform rank to every adapted weight matrix - a practical convenience that ignores a fundamental reality: different layers contribute unequally to task adaptation. We address this with a lightweight engineering solution: before fine-tuning begins, run eight calibration backward passes, compute the gradient variance of each LoRA-B matrix as a proxy for layer informativeness, and redistribute the rank budget proportionally. The resulting adapter is a standard LoRA with a per-layer rank pattern - no new parameters, no training overhead, no changes to serving infrastructure. We implement this via an efficient approximation of the empirical Fisher Information Matrix (eFIM) diagonal, restricted to LoRA adapter matrices only, which reduces memory cost by approximately 256x compared to full-model Fisher estimation. On GLUE with DeBERTa-v3-base, FIM-LoRA matches LoRA (88.6 vs. 88.7) at the same parameter budget, and on commonsense reasoning with LLaMA-3-8B reaches 68.5 vs. 68.7 for LoRA. The per-layer rank maps are interpretable: value projections and early-to-middle layers consistently receive higher rank, consistent with established findings on transformer layer roles.

Large language models (LLMs) have rapidly grown in scale, creating substantial memory and computational costs that hinder efficient deployment. Singular value decomposition (SVD) has emerged as an effective post-training compression technique, but existing SVD-based methods rely on static rank truncation, applying a fixed prefix of singular components to all inputs regardless of their diversity. We identify two limitations of this static design: the optimal rank varies across individual prompts, and the selected rank is sensitive to the choice of calibration set, leading to suboptimal performance across diverse inputs. To address these challenges, we propose $\textbf{PARSE}$, a post-training framework for $\textbf{P}$rompt-$\textbf{A}$ware $\textbf{R}$ank $\textbf{S}$election as $\textbf{E}$xperts in SVD-compressed LLMs. PARSE trains a linear router offline to perform prompt-aware rank selection, decoupling it from calibration information by supervising the router against dense-model outputs on a large-scale corpus. We further observe that rank-selection patterns are shared across semantically similar prompts and remain stable across decoding steps, allowing appropriate rank subsets to be served directly from a pattern cache at inference. Complemented by expert memory aggregation and kernel fusion for system-level efficiency, PARSE is orthogonal to existing SVD-based pipelines and consistently improves both model quality and inference efficiency. Integrated with four representative SVD-based methods, PARSE improves average task accuracy by up to 10% at a compression ratio of 0.6 on LLaMA-7B, and achieves up to 2.5 $\times$ prefill and 2.4 $\times$ decode speedup over native SVD execution.

Adaptive rank allocation for LoRA, allocating more parameters to important layers and fewer to unimportant ones, consistently improves efficiency under supervised fine-tuning (SFT). We investigate whether this success transfers to reinforcement learning, specifically Group Relative Policy Optimization (GRPO). Using gradient-magnitude profiling on Qwen 2.5 1.5B with GSM8K, we find that it does not: proportional rank allocation degrades accuracy by 4.5 points compared to uniform allocation (70.0% vs. 74.5%), despite using identical parameter budgets. We identify two mechanisms behind this failure. First, the gradient landscape under GRPO is fundamentally flatter than under SFT, the max-to-min layer importance ratio is only 2.17x, compared to >10x reported in SFT literature. All layers carry meaningful gradient signal; none are truly idle. Second, we discover a gradient amplification effect: non-uniform allocation widens the importance spread from 2.17x to 3.00x, creating a positive feedback loop where high-rank layers absorb more gradient while low-rank layers are progressively silenced. Our results suggest that gradient importance does not predict capacity requirements under RL, and that naive transfer of SFT-era rank allocation to alignment training should be avoided.

Generative Recommender (GR) inference places embedding hot caches (EMB) and KV caches in direct competition for limited GPU HBM: allocating more memory to one improves its efficiency but degrades the other. Existing systems optimize them in isolation, overlooking that the optimal EMB-KV allocation ratio can shift by up to 0.35 across workload regimes, leaving 20-30% latency improvement unrealized. While online reallocation is required to close this gap, naive approaches introduce H2D refill traffic on the critical path, causing P99 SLO violations. To address this, we present HELM, which jointly manages HBM allocation and request routing at runtime through two key components: (1) Adaptive Memory Allocation, a three-layer PPO-based controller (frozen base policy, online residual adapter, and burst-aware recovery controller) that achieves $32\,\mathrm{μs}$ decision latency while staying within 0.024-0.029 of the offline-optimal ratio; and (2) EMB-KV-Aware Scheduling, which routes requests by jointly considering KV residency, embedding locality, and node load to avoid routing inefficiencies under heterogeneous allocations. Evaluations on three production-scale datasets over a 32-node A100 cluster show that HELM reduces P99 latency by 24-38% over the best static policy and achieves 93.5-99.6% SLO satisfaction across Steady, Trend, and Burst workloads, significantly outperforming state-of-the-art baselines without sacrificing throughput.

Exponential growth in the scale of modern foundation models has led to the widespread adoption of Low-Rank Adaptation (LoRA) as a parameter-efficient fine-tuning technique. However, standard LoRA implementations disregard the varying intrinsic dimensionality of model layers and enforce a uniform rank, leading to parameter redundancy. We propose Post-Optimization Adaptive Rank Allocation (PARA), a data-free compression method for LoRA that integrates seamlessly into existing fine-tuning pipelines. PARA leverages Singular Value Decomposition to prune LoRA ranks using a global threshold over singular values across all layers. This results in non-uniform rank allocation based on layer-wise spectral importance. As a post-hoc method, PARA circumvents the training modifications and resulting instabilities that dynamic architectures typically incur. We empirically demonstrate that PARA reduces parameter count by 75-90% while preserving the predictive performance of the original, uncompressed LoRA across multiple vision and language benchmarks. Code will be published upon acceptance.

Mixture-of-Experts (MoE) has become a prominent paradigm for scaling Large Language Models (LLMs). Parameter-efficient fine-tuning methods, such as LoRA, are widely adopted to adapt pretrained MoE LLMs to downstream tasks. However, existing approaches typically assign identical LoRA ranks to all expert modules, ignoring the heterogeneous specialization of pretrained experts. This uniform allocation leads to a resource mismatch: task-relevant experts are under-provisioned, while less relevant ones receive redundant parameters. To address this, we propose DR-LoRA, a Dynamic Rank LoRA framework for fine-tuning pretrained MoE models. Specifically, DR-LoRA initializes all expert LoRA modules with a small active rank and uses an expert saliency score, which combines routing frequency and gradient-based rank importance, to identify which experts would benefit most from additional capacity. It then periodically expands the active ranks of the task-critical expert LoRA, progressively constructing a heterogeneous rank distribution tailored to the target task. Experiments on three MoE models across six tasks show that DR-LoRA consistently outperforms LoRA and other strong baselines, demonstrating that task-adaptive heterogeneous rank allocation is an effective strategy to improve active capacity utilization in MoE fine-tuning.