Revealing and Mitigating the Local Pattern Shortcuts of Mamba
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2024
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866913557230125056 |
|---|---|
| author | You, Wangjie Tang, Zecheng Li, Juntao Yao, Lili Zhang, Min |
| author_facet | You, Wangjie Tang, Zecheng Li, Juntao Yao, Lili Zhang, Min |
| contents | Large language models (LLMs) have advanced significantly due to the attention mechanism, but their quadratic complexity and linear memory demands limit their performance on long-context tasks. Recently, researchers introduced Mamba, an advanced model built upon State Space Models(SSMs) that offers linear complexity and constant memory. Although Mamba is reported to match or surpass the performance of attention-based models, our analysis reveals a performance gap: Mamba excels in tasks that involve localized key information but faces challenges with tasks that require handling distributed key information. Our controlled experiments suggest that this inconsistency arises from Mamba's reliance on local pattern shortcuts, which enable the model to remember local key information within its limited memory but hinder its ability to retain more dispersed information. Therefore, we introduce a global selection module into the Mamba model to address this issue. Experiments on both existing and proposed synthetic tasks, as well as real-world tasks, demonstrate the effectiveness of our method. Notably, with the introduction of only 4M extra parameters, our approach enables the Mamba model(130M) to achieve a significant improvement on tasks with distributed information, increasing its performance from 0 to 80.54 points. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2410_15678 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Revealing and Mitigating the Local Pattern Shortcuts of Mamba You, Wangjie Tang, Zecheng Li, Juntao Yao, Lili Zhang, Min Computation and Language Artificial Intelligence Large language models (LLMs) have advanced significantly due to the attention mechanism, but their quadratic complexity and linear memory demands limit their performance on long-context tasks. Recently, researchers introduced Mamba, an advanced model built upon State Space Models(SSMs) that offers linear complexity and constant memory. Although Mamba is reported to match or surpass the performance of attention-based models, our analysis reveals a performance gap: Mamba excels in tasks that involve localized key information but faces challenges with tasks that require handling distributed key information. Our controlled experiments suggest that this inconsistency arises from Mamba's reliance on local pattern shortcuts, which enable the model to remember local key information within its limited memory but hinder its ability to retain more dispersed information. Therefore, we introduce a global selection module into the Mamba model to address this issue. Experiments on both existing and proposed synthetic tasks, as well as real-world tasks, demonstrate the effectiveness of our method. Notably, with the introduction of only 4M extra parameters, our approach enables the Mamba model(130M) to achieve a significant improvement on tasks with distributed information, increasing its performance from 0 to 80.54 points. |
| title | Revealing and Mitigating the Local Pattern Shortcuts of Mamba |
| topic | Computation and Language Artificial Intelligence |
| url | https://arxiv.org/abs/2410.15678 |