Conditional probing: measuring usable information beyond a baseline
By John Hewitt, Kawin Ethayarajh, Percy Liang, Christopher D. Manning. 2021
68 citations recorded by the source corpus.
Probing experiments investigate the extent to which neural representations make properties—like part-of-speech—predictable. One suggests that a representation encodes a property if probing that representation produces higher accuracy than probing a baseline representation like non-contextual word embeddings. Instead of using baselines as a point of comparison, we’re interested in measuring information that is contained in the representation but not in the baseline. For example, current methods can detect when a representation is more useful than the word identity (a baseline) for predicting part-of-speech; however, they cannot detect when the representation is predictive of just the aspects of part-of-speech not explainable by the word identity. In this work, we extend a theory of usable information called V-information and propose conditional probing, which explicitly conditions on the information in the baseline. In a case study, we find that after conditioning on non-contextual word embeddings, properties like part-of-speech are accessible at deeper layers of a network than previously thought.
Subfields: Artificial Intelligence, Electrical and Electronic Engineering
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Artificial Intelligence
- Yejin Choi
- Xiang Ren
- Swabha Swayamdipta
- Kawin Ethayarajh
- Faeze Brahman
- Mohit Bansal
- Noah A. Smith
- Luke S. Zettlemoyer
In Electrical and Electronic Engineering
Citations in the corpus
- Understanding Dataset Difficulty with V-Usable Information (2021, neutral): “V -information (Xu et al., 2019) as well as to condition out information for probing-based interpretability techniques (Hewitt et al., 2021; Pimentel & Cotterell, 2021).”
- REV: Information-Theoretic Evaluation of Free-Text Rationales (2022, methodology): “Since CVI measures the additional V -usable information in R about Y beyond what’s already extracted from B by models in V , it grounds the goal of the proposed metric R EV .”
- Information-Theoretic Measures of Dataset Difficulty (2021, methodology): “In NLP, V - information has been used to study what context features Transformers actually use (O’Connor and Andreas, 2021), as well as to condition out information during probing (Hewitt et al., 2021), but not for estimating difficulty.”