Interpretability research should be judged by actionability—the degree to which its insights support concrete decisions and interventions—rather than explanatory power alone.
Title resolution pending
6 Pith papers cite this work. Polarity classification is still indexing.
6
Pith papers citing it
years
2026 6representative citing papers
Alignment in deep networks is governed by flag varieties, with subspace intersection dimension as the unique reparameterization-invariant observable, explaining regularization and activation effects from first principles.
Symmetric spectral diagnostics on attention are structurally blind to flow direction, with asymmetry G as the sole control parameter, yielding a two-axis test that distinguishes bottleneck versus diffuse hallucination modes with opposite polarity.
Sparse replacement layers decompose the MLP and attention modules of a chess-playing transformer to reveal verifiable tactical reasoning pathways and parallel computation patterns.
AP-MAE reconstructs masked attention patterns in LLMs with high accuracy, generalizes across models, predicts generation correctness at 55-70%, and enables 13.6% accuracy gains via targeted interventions.
CTT is a compression pipeline for LLMs that achieves up to 49x memory reduction, 10x faster inference, 81% lower CO2 emissions, and retains 68-98% accuracy on code clone detection, summarization, and generation tasks.
citing papers explorer
-
Interpretability Can Be Actionable
Interpretability research should be judged by actionability—the degree to which its insights support concrete decisions and interventions—rather than explanatory power alone.
-
Flag Varieties: A Geometric Framework for Deep Network Alignment
Alignment in deep networks is governed by flag varieties, with subspace intersection dimension as the unique reparameterization-invariant observable, explaining regularization and activation effects from first principles.
-
Self-Attention as Transport: Limits of Symmetric Spectral Diagnostics
Symmetric spectral diagnostics on attention are structurally blind to flow direction, with asymmetry G as the sole control parameter, yielding a two-axis test that distinguishes bottleneck versus diffuse hallucination modes with opposite polarity.
-
Tracing the Thought of a Grandmaster-level Chess-Playing Transformer
Sparse replacement layers decompose the MLP and attention modules of a chess-playing transformer to reveal verifiable tactical reasoning pathways and parallel computation patterns.
-
Automated Attention Pattern Discovery at Scale in Large Language Models
AP-MAE reconstructs masked attention patterns in LLMs with high accuracy, generalizes across models, predicts generation correctness at 55-70%, and enables 13.6% accuracy gains via targeted interventions.
-
Carbon-Taxed Transformers: A Green Compression Pipeline for Overgrown Language Models
CTT is a compression pipeline for LLMs that achieves up to 49x memory reduction, 10x faster inference, 81% lower CO2 emissions, and retains 68-98% accuracy on code clone detection, summarization, and generation tasks.