Prompt optimization per model substantially alters LLM rankings on both public and internal benchmarks compared to using fixed unoptimized prompts.
The LLM an- swer and the ground truth are compared using an LLM as a judge setup, with GPT-4o (Ope- nAI, 2024) utilized as the ‘judge’ model
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.AI 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Optimization before Evaluation: Evaluation with Unoptimised Prompts Can be Misleading
Prompt optimization per model substantially alters LLM rankings on both public and internal benchmarks compared to using fixed unoptimized prompts.