Investigating Tool-Memory Conflicts in Tool-Augmented LLMs
Fuente:
arXiv
Guardado en:
| Autores principales: | , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2026
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866909990766247936 |
|---|---|
| author | Cheng, Jiali Pan, Rui Amiri, Hadi |
| author_facet | Cheng, Jiali Pan, Rui Amiri, Hadi |
| contents | Tool-augmented large language models (LLMs) have powered many applications. However, they are likely to suffer from knowledge conflict. In this paper, we propose a new type of knowledge conflict -- Tool-Memory Conflict (TMC), where the internal parametric knowledge contradicts with the external tool knowledge for tool-augmented LLMs. We find that existing LLMs, though powerful, suffer from TMC, especially on STEM-related tasks. We also uncover that under different conditions, tool knowledge and parametric knowledge may be prioritized differently. We then evaluate existing conflict resolving techniques, including prompting-based and RAG-based methods. Results show that none of these approaches can effectively resolve tool-memory conflicts. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2601_09760 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Investigating Tool-Memory Conflicts in Tool-Augmented LLMs Cheng, Jiali Pan, Rui Amiri, Hadi Software Engineering Artificial Intelligence Tool-augmented large language models (LLMs) have powered many applications. However, they are likely to suffer from knowledge conflict. In this paper, we propose a new type of knowledge conflict -- Tool-Memory Conflict (TMC), where the internal parametric knowledge contradicts with the external tool knowledge for tool-augmented LLMs. We find that existing LLMs, though powerful, suffer from TMC, especially on STEM-related tasks. We also uncover that under different conditions, tool knowledge and parametric knowledge may be prioritized differently. We then evaluate existing conflict resolving techniques, including prompting-based and RAG-based methods. Results show that none of these approaches can effectively resolve tool-memory conflicts. |
| title | Investigating Tool-Memory Conflicts in Tool-Augmented LLMs |
| topic | Software Engineering Artificial Intelligence |
| url | https://arxiv.org/abs/2601.09760 |