Abstract
Attention-head ablation, zeroing a head and measuring the resulting change in task performance, is a common method for inferring which components of a language model are causally responsible for a behavior. We show using GPT-2 small that this inference can be fragile unless the intervention semantics, evaluation metric, and controls are carefully validated. A natural post-projection implementation of "zeroing a head" is nearly uncorrelated with a corrected pre-projection ablation (Pearson r = 0.057) and selects a completely disjoint top-5 set of important heads. We also show that binary accuracy can hide effects at behavioral floors and near ceilings, whereas gold-token log-probability remains graded. Using a discovery/held-out split and 1,000 matched random-head and layer-matched-head control draws, the corrected per-head effect ranking is highly stable across splits (Spearman rho = 0.974), and the top-5 selected heads significantly exceed both control distributions (Monte Carlo p = 0.001). However, evidence for task specificity is not robust on GPT-2. Replication on DistilGPT2 preserves the intervention-semantic and matched-control findings. These results show that single-head ablation does not by itself justify a causal claim; defensible interpretation requires correct intervention placement, a non-saturated continuous metric, and matched held-out controls.
Keywords
Subject
Publication details
- Journal
- Not available
- Open access
- Green open access
Cite this article
APA 7
Huang, J. (2026). When Do Attention-Head Ablations Support Causal Claims? Projection-Level Confounds, Floor Effects, and Matched Controls. https://omanscience.com/en/articles/when-do-attention-head-ablations-support-causal-claims-projection-level-confounds-floor-effects-and-matched-controls
MLA 9
Huang, Juli. "When Do Attention-Head Ablations Support Causal Claims? Projection-Level Confounds, Floor Effects, and Matched Controls." https://omanscience.com/en/articles/when-do-attention-head-ablations-support-causal-claims-projection-level-confounds-floor-effects-and-matched-controls.
Chicago (author–date)
Huang, Juli. 2026. "When Do Attention-Head Ablations Support Causal Claims? Projection-Level Confounds, Floor Effects, and Matched Controls." https://omanscience.com/en/articles/when-do-attention-head-ablations-support-causal-claims-projection-level-confounds-floor-effects-and-matched-controls.
Harvard
Huang, J. (2026) 'When Do Attention-Head Ablations Support Causal Claims? Projection-Level Confounds, Floor Effects, and Matched Controls', Available at: https://omanscience.com/en/articles/when-do-attention-head-ablations-support-causal-claims-projection-level-confounds-floor-effects-and-matched-controls.
Vancouver
Huang J. When Do Attention-Head Ablations Support Causal Claims? Projection-Level Confounds, Floor Effects, and Matched Controls. https://omanscience.com/en/articles/when-do-attention-head-ablations-support-causal-claims-projection-level-confounds-floor-effects-and-matched-controls
IEEE
J. Huang, "When Do Attention-Head Ablations Support Causal Claims? Projection-Level Confounds, Floor Effects, and Matched Controls," https://omanscience.com/en/articles/when-do-attention-head-ablations-support-causal-claims-projection-level-confounds-floor-effects-and-matched-controls.