frontier models
The Search Agent Keeps Evidence, Not Every Result
QRAE uses paired attention measurements to whitelist useful retrieved chunks while leaving uncertain evidence recoverable.
Summary
QRAE uses paired attention measurements to whitelist useful retrieved chunks while leaving uncertain evidence recoverable.
Query-Relevant Attention Entropy scores bounded batches with a query-free control and an entropy measure, then carries only high-scoring chunks into later reasoning. Uncertain chunks remain in a cache rather than being discarded. Across the reported DeepSearch benchmarks and models, QRAE improved task success by 2.35 percentage points on average while reducing latency by 30 percent and token use by 45 percent. The gains describe the tested systems, including one near 100 billion parameters.
Why it matters
QRAE uses paired attention measurements to whitelist useful retrieved chunks while leaving uncertain evidence recoverable.
Limits and context
- Query-Relevant Attention Entropy scores bounded batches with a query-free control and an entropy measure, then carries only high-scoring chunks into later reasoning.
- Uncertain chunks remain in a cache rather than being discarded.
Key claims
QRAE uses paired attention measurements to whitelist useful retrieved chunks while leaving uncertain evidence recoverable.
Qualification: Query-Relevant Attention Entropy scores bounded batches with a query-free control and an entropy measure, then carries only high-scoring chunks into later reasoning.
Evidence: source-2026-11-07-005
Sources
- arXiv preprint 2611.07976, original v1arXiv · primary research
Corrections
: Publication note: This November 7 edition was assembled before its scheduled 9:00 AM America/Detroit reveal but is being published late after recovery from a local build interruption. Its original reporting, image-generation times and scheduled publishAt are preserved.