A ReAct agent that reads a 13,572-line Ecore model via file tools matched direct full-context prompting on accuracy for the best models while using roughly 180 times fewer prompt tokens.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.SE 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Querying Large Automotive Software Models: Agentic vs. Direct LLM Approaches
A ReAct agent that reads a 13,572-line Ecore model via file tools matched direct full-context prompting on accuracy for the best models while using roughly 180 times fewer prompt tokens.