A new 15,000-sentence Marathi emotion benchmark shows GPT-4 and Llama3-405B outperform fine-tuned Marathi BERT and MuRIL, while BERT trained on GPT-4-generated labels still trails GPT-4.
Title resolution pending
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
-
L3Cube-MahaEmotions: A Marathi Emotion Recognition Dataset with Synthetic Annotations using CoTR prompting and Large Language Models
A new 15,000-sentence Marathi emotion benchmark shows GPT-4 and Llama3-405B outperform fine-tuned Marathi BERT and MuRIL, while BERT trained on GPT-4-generated labels still trails GPT-4.