Causal head-masking and dimension-zeroing experiments show retrieval heads are necessary for long-context recall and that low-frequency RoPE components within them drive performance across five models.
Gummadi, Manish Gupta, and Abhilasha Ravichander
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Does RoPE Prevent or Degrade Retrieval Heads? A Mechanistic Analysis Across Model Families
Causal head-masking and dimension-zeroing experiments show retrieval heads are necessary for long-context recall and that low-frequency RoPE components within them drive performance across five models.