ManySpace
Türkçe Sign in Sign up

AI & MLRecommendedAIML-RET-013

Training Dense Retrievers

Bi-encoders trained contrastively on query-passage pairs, with hard negatives doing most of the work.

Difficulty
4 / 5
Importance
4 / 5
Study time
1–2 h
Kind of work
Mixed

Why it matters

The learner can construct a contrastive training batch with in-batch and hard negatives and explain why hard negatives matter.

The route to this topic

Route AIML-RET-013
What Is a Model? Tasks: Regression, Classification, and Beyond Features and Targets Learning Paradigms: Where the Supervision Signal Comes From Classification Framing Linear Regression I: Prediction and Residuals Embeddings Tokens and Tokenization Scores and Logits Self-Supervised Learning Loss and Fitting: Squared Error Similarity in Vector Space Multiclass Scores and Softmax Objective Functions and Empirical Risk Semantic Embeddings Likelihood Cross-Entropy Loss Contrastive Learning Training Dense Retrievers