How we think about text classification in the LLM era
Medium explores the use of Large Language Models (LLMs) for text classification, particularly for updating their NSFW model, finding it effective but not suitable for all models due to cost and adaptability concerns.
MAIN POINTS
- Medium uses various ML models for story classification, affecting content recommendations.
- The outdated NSFW model needed revamping due to poor maintenance and performance.
- LLMs offer a simpler, training-free approach for text classification with customizable prompts.
- LLMs are less adaptive and more expensive, requiring careful consideration for use cases.
TAKEAWAYS
- LLMs can match or outperform legacy models in specific scenarios like NSFW classification.
- Cost estimation for LLMs is straightforward, but they can be expensive compared to custom models.
- LLMs are static and may not adapt to new trends, necessitating periodic updates.
- The NSFW model revamp was successful, but LLMs are not suitable for all Medium's models.