Masking-based explanations are query-channel communication: recovery fails above identification capacity and a sparse ML decoder succeeds below it, while LIME/SHAP-style surrogates lag.
Information dropout: Learning optimal representations through noisy computation
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.AI 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
The Query Channel: Information-Theoretic Limits of Masking-Based Explanations
Masking-based explanations are query-channel communication: recovery fails above identification capacity and a sparse ML decoder succeeds below it, while LIME/SHAP-style surrogates lag.