method
active
method:attention-sink-score

Attention Sink Score

Per-head metric measuring fraction of attention weight concentrated on first token position

Neighborhood — ranked by edge-count

Related by similarity (8)

cosine ≥ 0.65 · no typed edge

Entities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.

  • Attention Sinkconcept0.885
    Phenomenon where layers concentrate most attention weight on the first token, captured by ColSum Concentration metric
  • Attention streamsconcept0.758
    Term introduced by this paper for the horizontal axis of information flow in transformers across token-positions via attention heads
  • Sink Ratemethod0.753
    Fraction of attention heads with sink score above threshold 0.3, used to track stages of inference
  • Process using Q, K, V to compute a heat map over K and weighted sum of V.
  • Weighted Spearman correlation that corrects for sampling bias in automated interpretability evaluation
  • Probe scoreconcept0.733
    Dot product between hidden state and concept vector averaged across 5-layer window around best layer; measures model's internal emotive state
  • Attention Decayconcept0.729
    Decrease in attention paid to system prompt over conversational turns, leading to persona fidelity degradation (cited from Li et al.)
  • Attentionconcept0.727
    Selective attention mechanism is a component of GWT-2 indicator; AST-1 models and enables control over attention