Context-instrumental data distillation allows a 1.5B SLM to generate valid Kubernetes manifests at 91.5% pass@1 rate, with strict output formatting proving more impactful than additional training data.
arXiv:2402.10379 (2024)
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Context-Instrumental Data Distillation for Kubernetes Manifest Generation: Method and Experimental Evaluation
Context-instrumental data distillation allows a 1.5B SLM to generate valid Kubernetes manifests at 91.5% pass@1 rate, with strict output formatting proving more impactful than additional training data.