Efficient Mathematical Reasoning Models via Dynamic Pruning and Knowledge Distillation

• URL: http://arxiv.org/abs/2511.17577v1
• Date: Sat, 15 Nov 2025 09:21:44 GMT
• Title: Efficient Mathematical Reasoning Models via Dynamic Pruning and Knowledge Distillation
• Authors: Fengming Yu, Qingyu Meng, Haiwei Pan, Kejia Zhang,
• Abstract summary: This paper proposes a lightweight optimization method that integrates dynamic attention head pruning with knowledge distillation.<n>Experiments conducted on both Math23k and ASDiv-A verify the effectiveness of the proposed method.
• Score: 2.596115982322528
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: With the rapid development of deep learning, large language models have shown strong capabilities in complex reasoning tasks such as mathematical equation solving. However, their substantial computational and storage costs hinder practical deployment. This paper proposes a lightweight optimization method that integrates dynamic attention head pruning with knowledge distillation. The approach dynamically evaluates the importance of each attention head in the multi-head attention mechanism using a combination of weight norms and entropy, and prunes redundant heads in real time to reduce computational overhead. To mitigate performance degradation, knowledge distillation transfers information from the original model to the pruned student, enabling the smaller model to preserve reasoning ability. Experiments conducted on both Math23k and ASDiv-A verify the effectiveness of the proposed method. For example, on Math23k with a 30% pruning ratio, parameters are reduced by 18.7%, inference speed is improved by 27.5%, FLOPs are reduced by 19.3%, and accuracy drops only 0.7% (from 84.4% to 83.7%). These results demonstrate that the method achieves substantial efficiency gains while maintaining strong reasoning performance, providing a practical solution for efficient deployment of large language models in mathematical reasoning tasks.

Related papers

• HEFT: A Coarse-to-Fine Hierarchy for Enhancing the Efficiency and Accuracy of Language Model Reasoning [0.0]
  HEFT is a novel hierarchical adaptation strategy that composes two distinct PEFT methods in a coarse-to-fine manner.<n>A model fine-tuned for only three epochs with our HEFT strategy achieves an accuracy of 85.17%, exceeding the performance of models trained for 20 epochs.
  arXiv  Detail & Related papers  (2025-09-11T19:06:46Z)
• Efficient Machine Unlearning via Influence Approximation [75.31015485113993]
  Influence-based unlearning has emerged as a prominent approach to estimate the impact of individual training samples on model parameters without retraining.<n>This paper establishes a theoretical link between memorizing (incremental learning) and forgetting (unlearning)<n>We introduce the Influence Approximation Unlearning algorithm for efficient machine unlearning from the incremental perspective.
  arXiv  Detail & Related papers  (2025-07-31T05:34:27Z)
• Resource-Efficient Automatic Software Vulnerability Assessment via Knowledge Distillation and Particle Swarm Optimization [8.132644507041922]
  We propose a novel resource-efficient framework that integrates knowledge distillation and particle swarm optimization to enable automated vulnerability assessment.<n>Our framework employs a two-stage approach: First, particle swarm optimization is utilized to optimize the architecture of a compact student model.<n>Second, knowledge distillation is applied to transfer critical vulnerability assessment knowledge from a large teacher model to the optimized student model.
  arXiv  Detail & Related papers  (2025-07-30T13:55:28Z)
• Exploring and Exploiting the Inherent Efficiency within Large Reasoning Models for Self-Guided Efficiency Enhancement [101.77467538102924]
  Large reasoning models (LRMs) exhibit overthinking, which hinders efficiency and inflates inference cost.<n>We propose two lightweight methods to enhance LRM efficiency.<n>First, we introduce Efficiency Steering, a training-free activation steering technique that modulates reasoning behavior via a single direction.<n>Second, we develop Self-Rewarded Efficiency RL, a reinforcement learning framework that dynamically balances task accuracy and brevity.
  arXiv  Detail & Related papers  (2025-06-18T17:18:12Z)
• Think or Not? Exploring Thinking Efficiency in Large Reasoning Models via an Information-Theoretic Lens [51.90059610606049]
  This paper revisits the efficiency of such reasoning processes through an information-theoretic lens.<n>We propose two metrics, InfoBias and InfoGain, to quantify divergence from ideal reasoning paths and stepwise information contribution.<n>Motivated by these findings, we introduce an entropy-based Adaptive Think strategy that dynamically halts reasoning once confidence is sufficiently high.
  arXiv  Detail & Related papers  (2025-05-23T13:38:56Z)
• DRP: Distilled Reasoning Pruning with Skill-aware Step Decomposition for Efficient Large Reasoning Models [2.768827482823499]
  We propose Distilled Reasoning Pruning ( traces), a hybrid framework that combines inference-time pruning with tuning-based distillation.<n>We find that models trained with traces achieve substantial improvements in token efficiency without sacrificing accuracy.<n>Further analysis shows that aligning the reasoning structure of training CoTs with the student's reasoning capacity is critical for effective knowledge transfer and performance gains.
  arXiv  Detail & Related papers  (2025-05-20T06:15:15Z)
• Inference Scaling vs Reasoning: An Empirical Analysis of Compute-Optimal LLM Problem-Solving [0.0]
  Recent advances in large language models (LLMs) have predominantly focused on maximizing accuracy and reasoning capabilities.<n>This paper investigates the potential synergy between reasoning enhancement and computational efficiency by analyzing the integration of two contrasting approaches.
  arXiv  Detail & Related papers  (2024-12-20T08:42:45Z)
• Efficient Gravitational Wave Parameter Estimation via Knowledge Distillation: A ResNet1D-IAF Approach [2.4184866684341473]
  This study presents a novel approach using knowledge distillation techniques to enhance computational efficiency in gravitational wave analysis.<n>We develop a framework combining ResNet1D and Inverse Autoregressive Flow (IAF) architectures, where knowledge from a complex teacher model is transferred to a lighter student model.<n>Our experimental results show that the student model achieves a validation loss of 3.70 with optimal configuration (40,100,0.75), compared to the teacher model's 4.09, while reducing the number of parameters by 43%.
  arXiv  Detail & Related papers  (2024-12-11T03:56:46Z)
• Towards Compute-Optimal Transfer Learning [82.88829463290041]
  We argue that zero-shot structured pruning of pretrained models allows them to increase compute efficiency with minimal reduction in performance. Our results show that pruning convolutional filters of pretrained models can lead to more than 20% performance improvement in low computational regimes.
  arXiv  Detail & Related papers  (2023-04-25T21:49:09Z)
• Learning to Optimize Permutation Flow Shop Scheduling via Graph-based Imitation Learning [70.65666982566655]
  Permutation flow shop scheduling (PFSS) is widely used in manufacturing systems. We propose to train the model via expert-driven imitation learning, which accelerates convergence more stably and accurately. Our model's network parameters are reduced to only 37% of theirs, and the solution gap of our model towards the expert solutions decreases from 6.8% to 1.3% on average.
  arXiv  Detail & Related papers  (2022-10-31T09:46:26Z)
• Powerpropagation: A sparsity inducing weight reparameterisation [65.85142037667065]
  We introduce Powerpropagation, a new weight- parameterisation for neural networks that leads to inherently sparse models. Models trained in this manner exhibit similar performance, but have a distribution with markedly higher density at zero, allowing more parameters to be pruned safely. Here, we combine Powerpropagation with a traditional weight-pruning technique as well as recent state-of-the-art sparse-to-sparse algorithms, showing superior performance on the ImageNet benchmark.
  arXiv  Detail & Related papers  (2021-10-01T10:03:57Z)
• Towards Practical Lipreading with Distilled and Efficient Models [57.41253104365274]
  Lipreading has witnessed a lot of progress due to the resurgence of neural networks. Recent works have placed emphasis on aspects such as improving performance by finding the optimal architecture or improving generalization. There is still a significant gap between the current methodologies and the requirements for an effective deployment of lipreading in practical scenarios. We propose a series of innovations that significantly bridge that gap: first, we raise the state-of-the-art performance by a wide margin on LRW and LRW-1000 to 88.5% and 46.6%, respectively using self-distillation.
  arXiv  Detail & Related papers  (2020-07-13T16:56:27Z)

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.