SAGE is a training-free context reduction method that converts attention signals from a small LLM into a differential relevance heatmap to select top units for downstream QA, achieving competitive accuracy at 10% token budget on benchmarks like QuALITY-hard.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.DB 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
SAGE: Selective Attention-Guided Extraction for Token-Efficient Document Indexing
SAGE is a training-free context reduction method that converts attention signals from a small LLM into a differential relevance heatmap to select top units for downstream QA, achieving competitive accuracy at 10% token budget on benchmarks like QuALITY-hard.