Semi-Synthetic Transformers for Evaluating Mechanistic Interpretability Techniques

Oct-10-2025, 12:42:16 GMT–Neural Information Processing Systems

Training (IIT) which we call Strict IIT (SIIT). SIIT models maintain Tracr's original circuit while being more realistic.

node, sequence, transformer, (12 more...)

Neural Information Processing Systems

Oct-10-2025, 12:42:16 GMT

Conferences PDF

Country:
- Asia > Middle East
  - Jordan (0.04)
- South America > Argentina
  - Pampas > Buenos Aires F.D. > Buenos Aires (0.04)

Genre:
- Research Report
  - Experimental Study (0.68)
  - New Finding (0.93)

Technology:
- Information Technology
  - Artificial Intelligence
    - Machine Learning
      - Neural Networks (0.69)
      - Performance Analysis > Accuracy (0.93)
    - Natural Language (1.00)
  - Data Science (0.68)

Duplicate Docs Excel Report

Title
a8f7d43ae092d9a5295775eb17f3f4f7-Supplemental-Datasets_and_Benchmarks_Track.pdf
a8f7d43ae092d9a5295775eb17f3f4f7-Supplemental-Datasets_and_Benchmarks_Track.pdf
Semi-Synthetic Transformers for Evaluating Mechanistic Interpretability Techniques

Similar Docs Excel Report more

Title	Similarity	Source
None found