pith. sign in

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 1

years

2026 1

verdicts

UNVERDICTED 1

representative citing papers

Explaining Attention with Program Synthesis

cs.LG · 2026-06-17 · unverdicted · novelty 7.0

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.

citing papers explorer

Showing 1 of 1 citing paper.

  • Explaining Attention with Program Synthesis cs.LG · 2026-06-17 · unverdicted · none · ref 20

    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.