Sparse Feature Circuits: Discovering and Editing Interpretable Causal Graphs in Language Models.
Samuel MarksCan RagerEric J. MichaudYonatan BelinkovDavid BauAaron MuellerPublished in: CoRR (2024)
Keyphrases
- language model
- causal graph
- language modeling
- n gram
- document retrieval
- probabilistic model
- speech recognition
- query expansion
- information retrieval
- statistical language models
- retrieval model
- test collection
- smoothing methods
- context sensitive
- ad hoc information retrieval
- language modelling
- feature vectors
- language models for information retrieval
- pseudo relevance feedback
- planning problems
- vector space model
- relevance model
- document ranking
- word segmentation
- language model for information retrieval
- statistical language modeling
- feature selection
- information retrieval systems
- translation model
- query terms
- causal models
- state variables
- domain independent
- machine learning
- retrieval effectiveness