ManySpace
Türkçe Sign in Sign up

AI & MLRecommendedAIML-INT-004

Probing Representations

Training simple classifiers on hidden states to test what information a layer encodes.

Difficulty
3 / 5
Importance
4 / 5
Study time
1–2 h
Kind of work
Mixed

Why it matters

The learner can train a linear probe across layers and explain why high probe accuracy does not show the model uses that information.

The route to this topic

Route AIML-INT-004
What Is a Model? Tasks: Regression, Classification, and Beyond Features and Targets Classification Framing Numeric and Categorical Data Linear Regression I: Prediction and Residuals Embeddings Scores and Logits Encoding Categories Loss and Fitting: Squared Error Similarity in Vector Space Sigmoid and Probabilities Distributed Representations Logistic Regression Word Embeddings Contextual Representations Probing Representations