AutoEvoEval applies 22 atomic perturbations and multi-round chains to MCQ benchmarks, causing average accuracy drops of 7.283% and up to 52.932% for long chains.
Deepseek-v3: Scaling open-source language models with mixture of experts
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
AutoEvoEval: An Automated Framework for Evolving Close-Ended LLM Evaluation Data
AutoEvoEval applies 22 atomic perturbations and multi-round chains to MCQ benchmarks, causing average accuracy drops of 7.283% and up to 52.932% for long chains.