Stingy Context: 18:1 Hierarchical Code Compression for LLM Auto-Coding
Fuente:
arXiv
Guardado en:
| Autor principal: | |
|---|---|
| Formato: | Preprint |
| Publicado: |
2026
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866911421654106112 |
|---|---|
| author | Ostby, David Linus |
| author_facet | Ostby, David Linus |
| contents | We introduce Stingy Context, a hierarchical tree-based compression scheme achieving 18:1 reduction in LLM context for auto-coding tasks. Using our TREEFRAG exploit decomposition, we reduce a real source code base of 239k tokens to 11k tokens while preserving task fidelity. Empirical results across 12 Frontier models show 94 to 97% success on 40 real-world issues at low cost, outperforming flat methods and mitigating lost-in-the-middle effects. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2601_19929 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Stingy Context: 18:1 Hierarchical Code Compression for LLM Auto-Coding Ostby, David Linus Computation and Language Artificial Intelligence Software Engineering We introduce Stingy Context, a hierarchical tree-based compression scheme achieving 18:1 reduction in LLM context for auto-coding tasks. Using our TREEFRAG exploit decomposition, we reduce a real source code base of 239k tokens to 11k tokens while preserving task fidelity. Empirical results across 12 Frontier models show 94 to 97% success on 40 real-world issues at low cost, outperforming flat methods and mitigating lost-in-the-middle effects. |
| title | Stingy Context: 18:1 Hierarchical Code Compression for LLM Auto-Coding |
| topic | Computation and Language Artificial Intelligence Software Engineering |
| url | https://arxiv.org/abs/2601.19929 |