Disrupting the concentrated massive values in Q and K of RoPE-based LLMs collapses contextual understanding tasks while leaving parametric retrieval mostly intact, and the paper attributes the pattern to RoPE.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Massive Values in Self-Attention Modules are the Key to Contextual Knowledge Understanding
Disrupting the concentrated massive values in Q and K of RoPE-based LLMs collapses contextual understanding tasks while leaving parametric retrieval mostly intact, and the paper attributes the pattern to RoPE.