Context Discipline and Performance Correlation: Analyzing LLM Performance and Quality Degradation Under Varying Context Lengths
Fuente:
arXiv
Guardado en:
| Autores principales: | , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2025
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866917207272849408 |
|---|---|
| author | Ponnusamy, Ahilan Ayyachamy Nadar Chandran, Karthic Hossain, M Maruf |
| author_facet | Ponnusamy, Ahilan Ayyachamy Nadar Chandran, Karthic Hossain, M Maruf |
| contents | The scaling trend in Large Language Models (LLMs) has prioritized increasing the maximum context window to facilitate complex, long-form reasoning and document analysis. However, managing this expanded context introduces severe computational overhead. This paper investigates the critical trade-off between system performance and model quality when dense transformer architectures--specifically Llama-3.1-70B and Qwen1.5-14B--are exposed to large volumes of irrelevant and distracting context. The research identifies a non-linear performance degradation tied to the growth of the Key-Value (KV) cache. Furthermore, an extended analysis of the Mixture-of-Experts (MoE) architecture reveals unique behavioral anomalies at varying context scales, suggesting that architectural benefits may be masked by infrastructure bottlenecks at high token volumes. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2601_11564 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Context Discipline and Performance Correlation: Analyzing LLM Performance and Quality Degradation Under Varying Context Lengths Ponnusamy, Ahilan Ayyachamy Nadar Chandran, Karthic Hossain, M Maruf Computation and Language Artificial Intelligence The scaling trend in Large Language Models (LLMs) has prioritized increasing the maximum context window to facilitate complex, long-form reasoning and document analysis. However, managing this expanded context introduces severe computational overhead. This paper investigates the critical trade-off between system performance and model quality when dense transformer architectures--specifically Llama-3.1-70B and Qwen1.5-14B--are exposed to large volumes of irrelevant and distracting context. The research identifies a non-linear performance degradation tied to the growth of the Key-Value (KV) cache. Furthermore, an extended analysis of the Mixture-of-Experts (MoE) architecture reveals unique behavioral anomalies at varying context scales, suggesting that architectural benefits may be masked by infrastructure bottlenecks at high token volumes. |
| title | Context Discipline and Performance Correlation: Analyzing LLM Performance and Quality Degradation Under Varying Context Lengths |
| topic | Computation and Language Artificial Intelligence |
| url | https://arxiv.org/abs/2601.11564 |