Enregistré dans:
| Auteurs principaux: | , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2025
|
| Sujets: | |
| Accès en ligne: | https://arxiv.org/abs/2507.03828 |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866913048426446848 |
|---|---|
| author | Chowdhury, Md Mokarram Asante, Daniel Agyei Chang, Ernie Li, Yang |
| author_facet | Chowdhury, Md Mokarram Asante, Daniel Agyei Chang, Ernie Li, Yang |
| contents | Large language models (LLMs) achieve strong performance across diverse domains but remain difficult to deploy in resource-constrained environments due to their size. Low-rank compression is a common remedy, typically minimizing weight reconstruction error under the assumption that weights are low-rank. However, this assumption often does not hold in LLMs. In contrast, LLM activations exhibit a more pronounced low-rank structure, motivating approaches that minimize activation reconstruction error.
This shift alone, however, is not sufficient: different activation dimensions contribute unequally to model performance, and treating them uniformly can lead to accuracy loss. We introduce IMPACT, an importance-aware activation reconstruction framework that links compression to its effect on model performance. IMPACT formulates compression as an optimization problem that integrates activation structure with gradient-based importance, deriving a closed-form solution where reconstruction bases arise from an importance-weighted activation covariance matrix. This yields low-rank compression explicitly optimized for accuracy preservation. Experiments across multiple models and tasks demonstrate that IMPACT achieves up to 55.4% greater model size reduction while maintaining accuracy comparable to or better than state-of-the-art baselines. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2507_03828 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | IMPACT: Importance-Aware Activation Space Reconstruction Chowdhury, Md Mokarram Asante, Daniel Agyei Chang, Ernie Li, Yang Machine Learning Large language models (LLMs) achieve strong performance across diverse domains but remain difficult to deploy in resource-constrained environments due to their size. Low-rank compression is a common remedy, typically minimizing weight reconstruction error under the assumption that weights are low-rank. However, this assumption often does not hold in LLMs. In contrast, LLM activations exhibit a more pronounced low-rank structure, motivating approaches that minimize activation reconstruction error. This shift alone, however, is not sufficient: different activation dimensions contribute unequally to model performance, and treating them uniformly can lead to accuracy loss. We introduce IMPACT, an importance-aware activation reconstruction framework that links compression to its effect on model performance. IMPACT formulates compression as an optimization problem that integrates activation structure with gradient-based importance, deriving a closed-form solution where reconstruction bases arise from an importance-weighted activation covariance matrix. This yields low-rank compression explicitly optimized for accuracy preservation. Experiments across multiple models and tasks demonstrate that IMPACT achieves up to 55.4% greater model size reduction while maintaining accuracy comparable to or better than state-of-the-art baselines. |
| title | IMPACT: Importance-Aware Activation Space Reconstruction |
| topic | Machine Learning |
| url | https://arxiv.org/abs/2507.03828 |