Low-Rank Compression of Language Models via Differentiable Rank Selection

• URL: http://arxiv.org/abs/2512.13733v1
• Date: Sun, 14 Dec 2025 07:20:57 GMT
• Title: Low-Rank Compression of Language Models via Differentiable Rank Selection
• Authors: Sidhant Sundrani, Francesco Tudisco, Pasquale Minervini,
• Abstract summary: We propose Learning to Low-Rank Compress (LLRC), a gradient-based approach which directly learns the weights of masks that select singular values in a fine-tuning-free setting.<n>Our approach outperforms competing ranking selection methods that similarly require no post-compression fine-tuning across various compression rates on common-sense reasoning and open-domain question-answering tasks.
• Score: 22.99526059495007
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Approaches for compressing large-language models using low-rank decomposition have made strides, particularly with the introduction of activation and loss-aware SVD, which improves the trade-off between decomposition rank and downstream task performance. Despite these advancements, a persistent challenge remains--selecting the optimal ranks for each layer to jointly optimise compression rate and downstream task accuracy. Current methods either rely on heuristics that can yield sub-optimal results due to their limited discrete search space or are gradient-based but are not as performant as heuristic approaches without post-compression fine-tuning. To address these issues, we propose Learning to Low-Rank Compress (LLRC), a gradient-based approach which directly learns the weights of masks that select singular values in a fine-tuning-free setting. Using a calibration dataset, we train only the mask weights to select fewer and fewer singular values while minimising the divergence of intermediate activations from the original model. Our approach outperforms competing ranking selection methods that similarly require no post-compression fine-tuning across various compression rates on common-sense reasoning and open-domain question-answering tasks. For instance, with a compression rate of 20% on Llama-2-13B, LLRC outperforms the competitive Sensitivity-based Truncation Rank Searching (STRS) on MMLU, BoolQ, and OpenbookQA by 12%, 3.5%, and 4.4%, respectively. Compared to other compression techniques, our approach consistently outperforms fine-tuning-free variants of SVD-LLM and LLM-Pruner across datasets and compression rates. Our fine-tuning-free approach also performs competitively with the fine-tuning variant of LLM-Pruner.

Related papers

• Zero Sum SVD: Balancing Loss Sensitivity for Low Rank LLM Compression [11.908793753919745]
  We propose textbfZero Sum SVD (textbfZS-SVD), a post-training method that performs singular component selection in whitened coordinates.<n>textbfZS-SVD prunes components across the whole model with a textbfzero sum rule that keeps the cumulative predicted loss change near zero.<n>Experiments show consistent gains across diverse benchmarks and compression ratios.
  arXiv  Detail & Related papers  (2026-02-02T21:51:01Z)
• SkipCat: Rank-Maximized Low-Rank Compression of Large Language Models via Shared Projection and Block Skipping [6.789200833454491]
  Large language models (LLM) have achieved remarkable performance across a wide range of tasks.<n>Low-rank compression is a promising approach to address this issue, as it reduces both computational and memory costs.<n>We propose SkipCat, a novel low-rank compression framework that enables the use of higher ranks while achieving the same compression rates.
  arXiv  Detail & Related papers  (2025-12-15T16:25:55Z)
• Rethinking LLM Evaluation: Can We Evaluate LLMs with 200x Less Data? [82.09573568241724]
  EssenceBench is a coarse-to-fine framework utilizing an iterative Genetic Algorithm (GA)<n>Our approach yields superior compression results with lower reconstruction error and markedly higher efficiency.<n>On the HellaSwag benchmark (10K samples), our method preserves the ranking of all models shifting within 5% using 25x fewer samples, and achieves 95% ranking preservation shifting within 5% using only 200x fewer samples.
  arXiv  Detail & Related papers  (2025-10-12T05:38:10Z)
• Greedy Low-Rank Gradient Compression for Distributed Learning with Convergence Guarantees [10.828702910680692]
  We propose the first Greedy Low-Rank compression algorithm for distributed learning with rigorous convergence guarantees.<n>We prove that GreedyLore achieves a convergence rate of $mathcalO(sigma/sqrtNT + 1/T)$ under standards such as MSGD and Adam--marking the first linear speedup convergence rate for low-rank gradient compression.
  arXiv  Detail & Related papers  (2025-07-11T17:46:12Z)
• MGAA: Multi-Granular Adaptive Allocation fof Low-Rank Compression of LLMs [9.244526043014098]
  Multi-Granular Adaptive Allocation (MGAA) method can adaptively allocate parameters between and within sublayers without task-specific evaluations in the compression process.<n> Comprehensive evaluations of MGAA across multiple LLMs backbone models and benchmark datasets demonstrate its superior performance.
  arXiv  Detail & Related papers  (2025-07-04T04:54:01Z)
• Choose Your Model Size: Any Compression of Large Language Models Without Re-Computation [10.376875638696504]
  This work presents Any Compression via Iterative Pruning (ACIP), a novel algorithmic approach to determine a compression-performance trade-off.<n>We use an SVD-reparametrization of linear layers and iteratively prune their singular values with a sparsity-inducing penalty.<n>We show that ACIP seamlessly complements common quantization-based compression techniques.
  arXiv  Detail & Related papers  (2025-02-03T18:40:58Z)
• GRASP: Replace Redundant Layers with Adaptive Singular Parameters for Efficient Model Compression [26.51079570548107]
  We propose GRASP (Gradient-based Retention of Adaptive Singular Parameters), a novel compression framework.<n>By replacing redundant layers with only a minimal set of parameters, GRASP achieves efficient compression while maintaining strong performance with minimal overhead.
  arXiv  Detail & Related papers  (2024-12-31T08:22:21Z)
• ALoRE: Efficient Visual Adaptation via Aggregating Low Rank Experts [71.91042186338163]
  ALoRE is a novel PETL method that reuses the hypercomplex parameterized space constructed by Kronecker product to Aggregate Low Rank Experts.<n>Thanks to the artful design, ALoRE maintains negligible extra parameters and can be effortlessly merged into the frozen backbone.
  arXiv  Detail & Related papers  (2024-12-11T12:31:30Z)
• EoRA: Fine-tuning-free Compensation for Compressed LLM with Eigenspace Low-Rank Approximation [84.70637613266835]
  EoRA is a fine-tuning-free method that augments compressed Large Language Models with low-rank matrices.<n>EoRA consistently outperforms prior training-free low rank methods in recovering the accuracy of compressed LLMs.
  arXiv  Detail & Related papers  (2024-10-28T17:59:03Z)
• LoRC: Low-Rank Compression for LLMs KV Cache with a Progressive Compression Strategy [59.1298692559785]
  Key-Value ( KV) cache is crucial component in serving transformer-based autoregressive large language models (LLMs) Existing approaches to mitigate this issue include: (1) efficient attention variants integrated in upcycling stages; (2) KV cache compression at test time; and (3) KV cache compression at test time. We propose a low-rank approximation of KV weight matrices, allowing plug-in integration with existing transformer-based LLMs without model retraining. Our method is designed to function without model tuning in upcycling stages or task-specific profiling in test stages.
  arXiv  Detail & Related papers  (2024-10-04T03:10:53Z)
• Flattened one-bit stochastic gradient descent: compressed distributed optimization with controlled variance [55.01966743652196]
  We propose a novel algorithm for distributed gradient descent (SGD) with compressed gradient communication in the parameter-server framework. Our gradient compression technique, named flattened one-bit gradient descent (FO-SGD), relies on two simple algorithmic ideas.
  arXiv  Detail & Related papers  (2024-05-17T21:17:27Z)
• ROPO: Robust Preference Optimization for Large Language Models [59.10763211091664]
  We propose an iterative alignment approach that integrates noise-tolerance and filtering of noisy samples without the aid of external models. Experiments on three widely-used datasets with Mistral-7B and Llama-2-7B demonstrate that ROPO significantly outperforms existing preference alignment methods.
  arXiv  Detail & Related papers  (2024-04-05T13:58:51Z)
• Sparse is Enough in Fine-tuning Pre-trained Large Language Models [98.46493578509039]
  We propose a gradient-based sparse fine-tuning algorithm, named Sparse Increment Fine-Tuning (SIFT) We validate its effectiveness on a range of tasks including the GLUE Benchmark and Instruction-tuning.
  arXiv  Detail & Related papers  (2023-12-19T06:06:30Z)
• Learning Accurate Performance Predictors for Ultrafast Automated Model Compression [86.22294249097203]
  We propose an ultrafast automated model compression framework called SeerNet for flexible network deployment. Our method achieves competitive accuracy-complexity trade-offs with significant reduction of the search cost.
  arXiv  Detail & Related papers  (2023-04-13T10:52:49Z)
• Attentional-Biased Stochastic Gradient Descent [74.49926199036481]
  We present a provable method (named ABSGD) for addressing the data imbalance or label noise problem in deep learning. Our method is a simple modification to momentum SGD where we assign an individual importance weight to each sample in the mini-batch. ABSGD is flexible enough to combine with other robust losses without any additional cost.
  arXiv  Detail & Related papers  (2020-12-13T03:41:52Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.