Natural Language Inference On Lidirus

评估指标

MCC

评测结果

各个模型在此基准测试上的表现结果

Paper TitleRepository
Human Benchmark0.626RussianSuperGLUE: A Russian Language Understanding Evaluation Benchmark
ruRoberta-large finetune0.339--
ruT5-large-finetune0.32--
ruT5-base-finetune0.267--
ruBert-large finetune0.235--
RuGPT3Large0.231--
ruBert-base finetune0.224--
SBERT_Large_mt_ru_finetuning0.218--
SBERT_Large0.209--
RuBERT plain0.191--
Multilingual Bert0.189--
RuBERT conversational0.178--
heuristic majority0.147Unreasonable Effectiveness of Rule-Based Heuristics in Solving Russian SuperGLUE Tasks-
YaLM 1.0B few-shot0.124--
RuGPT3XL few-shot0.096--
MT5 Large0.061mT5: A massively multilingual pre-trained text-to-text transformer
Baseline TF-IDF1.10.06RussianSuperGLUE: A Russian Language Understanding Evaluation Benchmark
RuGPT3Medium0.01--
majority_class0Unreasonable Effectiveness of Rule-Based Heuristics in Solving Russian SuperGLUE Tasks-
Random weighted0Unreasonable Effectiveness of Rule-Based Heuristics in Solving Russian SuperGLUE Tasks-
0 of 22 row(s) selected.
Natural Language Inference On Lidirus | SOTA | HyperAI超神经