Context compression improves search efficiency and accuracy

上下文压缩技术提升搜索效率与准确性

Perplexity has deployed query-aware compression in production, reducing context tokens by up to 70% while improving search answer quality.

Why it matters: HKR-H/K/R all pass: a counterintuitive production search update with a 70% token-cut claim and a direct cost-latency-quality hook. Single-source X post lacks benchmarks and reproducible setup, so it stays in the lower good-quality band.