notesum.ai

Published at November 11

1-800-SHARED-TASKS @ NLU of Devanagari Script Languages: Detection of Language, Hate Speech, and Targets using LLMs

cs.CL

cs.AI

cs.LG

Released Date: November 11, 2024

Authors: Jebish Purbey¹, Siddartha Pullakhandam², Kanwal Mehreen³, Muhammad Arham⁴, Drishti Sharma⁵, Ashay Srivastava⁶, Ram Mohan Rao Kadiyala⁷

Aff.: ¹Pulchowk Campus, IoE; ²University of Wisconsin; ³Traversaal.ai; ⁴NUST/Traversaal.ai; ⁵Cohere For AI Community; ⁶University of Maryland; ⁷University of Maryland, College Park

Arxiv: http://arxiv.org/abs/2411.06850v1

Model

Description

MuRIL

Fine-tuned on train+dev set

0.9968

IndicBERT V2

Fine-tuned on train+dev set

0.9977

Gemma-2 9B

Fine-tuned on train+dev set

0.9973

Ensemble-1

MuRIL’s prediction as fallback

in case of no majority

0.9979

Ensemble-2

IndicBERT V2’s prediction as

fallback in case of no majority

0.9980

Ensemble-3

Gemma-2 9B’s prediction as

fallback in case of no majority

0.9979