SPEAR augments automatic prompt engineering with an agent that uses evaluate, python, set_prompt and finish tools plus auto-rollback guardrails, outperforming baselines on industrial LLM-judge tasks and BBH by leveraging Python-authored structural error analysis.
InProceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Vol- ume 1: Long Papers), pages 1658–1677, Bangkok, Thailand
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
SPEAR: Code-Augmented Agentic Prompt Optimization
SPEAR augments automatic prompt engineering with an agent that uses evaluate, python, set_prompt and finish tools plus auto-rollback guardrails, outperforming baselines on industrial LLM-judge tasks and BBH by leveraging Python-authored structural error analysis.