Summary of Self-supervised Interpretable Concept-based Models For Text Classification, by Francesco De Santis et al.
Self-supervised Interpretable Concept-based Models for Text Classification
by Francesco De Santis, Philippe Bich, Gabriele Ciravegna, Pietro Barbiero, Danilo Giordano, Tania Cerquitelli
First submitted to arxiv on: 20 Jun 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Artificial Intelligence (cs.AI)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary The proposed Interpretable Concept Embedding Models (ICEMs) aim to overcome the limitations of Large-Language Models (LLMs) by providing explainable-by-design architectures for textual data. By leveraging the generalization abilities of LLMs, ICEMs predict concept labels in a self-supervised manner while delivering final predictions with an interpretable function. The models achieve similar performance to fully supervised and end-to-end black-box models, offering logical explanations for their predictions. Additionally, ICEMs are interactable, allowing humans to modify intermediate predictions through concept interventions, and controllable, guiding the LLMs’ decoding process. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper proposes a new way to make Large-Language Models (LLMs) more understandable. Instead of using traditional methods that try to explain how the model works, the authors create a new type of model called Interpretable Concept Embedding Models (ICEMs). These models can predict what’s going on inside the LLM and why it made certain decisions. This helps make the LLM more reliable and easier to use. |
Keywords
» Artificial intelligence » Embedding » Generalization » Self supervised » Supervised
