In-context learning strategies for political bias detection (text classification) using LlaMa-3-70b

What is it about?

We used the dataset, MBIB (Media Bias Identification Benchmark Dataset), which comprises various types of bias splits. This political bias detection dataset was used in this study. We tested the performance of Meta AI's latest LLM, LLaMa-3-70b (70 billion model) using zero, few- shot and chain-of-thought prompting. In the chain of thought prompt, we provided some subtle, nuanced reasons. These reasons also depict the complexity of the political bias detection in media and news text. The chain-of-thought prompt was able to enhance the performance of the model to a significant level. Showing potential in such strategies for bias detection.

Featured Image