# Perplexity trains embedding models on distilled chunk-level relevance scores

> Original title: The models are trained differently from other contextual embeddings models.

- Source: Perplexity Developers
- Published: 2026-11-02T21:50:58.000Z
- Bad Influence: https://badinfluence.news/items/ma8omeig8tl4tlbk4i573crsr
- Original: https://x.com/perplexitydevs/status/2117373931703455908

## Summary

Perplexity says its contextual embedding models are trained differently from other contextual embeddings models. Instead of one gold chunk per query, chunk-level relevance scores are distilled from a query-aware context compression model in its answer pipeline, and supporting chunks are no longer treated as negatives.

## Post

The models are trained differently from other contextual embeddings models.

Instead of one gold chunk per query, chunk-level relevance scores are distilled from a query-aware context compression model in Perplexity's answer pipeline. Supporting chunks are no longer negatives.

![](https://badinfluence.news/api/img-proxy?u=https%3A%2F%2Fpbs.twimg.com%2Fmedia%2FHWJsXuMbwAAtv2t.png&mode=thumb&exp=1793923200&sig=1dca27583d3c3bef)