How we think about text classification in the LLM era

Summary

Medium revisited its aging NSFW text-classification model, which had outdated code, infrequent retraining, and poorly documented training and performance history. The replacement uses an LLM prompt containing NSFW guidelines and story content, requesting a 0-to-100 score, structured output, and a short reason rather than requiring a training set. To evaluate it, the team created a curator-labeled dataset with deliberate oversampling of NSFW stories, then compared several LLMs with the legacy model using false-positive and false-negative metrics. Some models matched or exceeded the legacy model; the team selected Mixtral 8×7b based on cost and performance and ran inference in Snowflake. The post presents this as a strong fit where training data is costly, trends are relatively stable, and only distribution-eligible stories need scoring, while topic and spam models still benefit from continuously updated datasets.

Takeaways

1. 1

   An LLM-based NSFW classifier matched or outperformed the legacy model on false positives and false negatives in the team’s evaluation.

2. 2

   Requesting a score instead of only yes/no enables ranking, ROC-AUC or PR-AUC evaluation, and later threshold changes to adjust the false-positive and false-negative balance.

3. 3

   The LLM approach fit NSFW classification because maintaining labeled data was costly and trends were judged relatively stable, but topic and spam models still benefit from continuously refreshed training data.