Language-model-guided program synthesis can approximate transformer attention heads with over 75% IoU fidelity on held-out data and allow replacing 25% of heads with only 16% average perplexity increase.
Can interpretation predict behavior on unseen data?arXiv preprint arXiv:2507.06445, 2025
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Explaining Attention with Program Synthesis
Language-model-guided program synthesis can approximate transformer attention heads with over 75% IoU fidelity on held-out data and allow replacing 25% of heads with only 16% average perplexity increase.