2023 · 99 citations · 25 references
ComputingComputer ScienceComputational LinguisticsFeature EngineeringNatural Language ProcessingEnsemble MethodsLanguage ModelingText GenerationText SummarizationBenchmark Dataset DesignBenchmark Dataset DevelopmentBenchmark Dataset EvaluationLarge Language ModelsMetricsPairwise ComparisonPairwise RankingPairwise Ranking Models
We present LLM-Blender, an ensembling framework designed to attain consistently superior performance by leveraging the diverse strengths of multiple open-source large language models (LLMs). Our framework consists of two modules: PairRanker and GenFuser, addressing the observation that optimal LLMs for different examples can significantly vary. PairRanker employs a specialized pairwise comparison method to distinguish subtle differences between candidate outputs. It jointly encodes the input text and a pair of candidates, using cross-attention encoders to determine the superior one. Our results demonstrate that PairRanker exhibits the highest correlation with ChatGPT-based ranking. Then, GenFuser aims to merge the top-ranked candidates, generating an improved output by capitalizing on their strengths and mitigating their weaknesses. To facilitate large-scale evaluation, we introduce a benchmark dataset, MixInstruct, which is a mixture of multiple instruction datasets featuring oracle pairwise comparisons. Our LLM-Blender significantly outperform individual LLMs and baseline methods across various metrics, establishing a substantial performance gap.
25
Training language models to follow instructions with human feedback
Long Ouyang, Jeff Wu, Xu Jiang et al. · arXiv (Cornell University) · 2022 · 4.3K citations · Full text
LLaMA: Open and Efficient Foundation Language Models
Hugo Touvron, Thibaut Lavril, Gautier Izacard et al. · arXiv (Cornell University) · 2023 · 3.9K citations · Full text