KS Research LabAdvanced Engineering & AI
EMPIRICAL BENCHMARKS

Benchmarks Explorer

Empirical measurements, comparative latency, and throughput baselines. We publish verifiable figures — never synthetic marketing counters or fabricated metrics.

BM-002RS-003

Zero-Copy Linear Memory Ingestion Throughput

Edge Worker V8 / Wasm32 Contiguous Memory
Baseline Reference:V8 JSON.parse + dynamic object allocation
KS Lab Candidate:Wasm Linear Memory Zero-Copy Buffer Mapping
Payload Ingestion Latency
0.12ms
LOWER IS BETTER
Heap Memory Allocation
4.2MB
LOWER IS BETTER
Ingestion Throughput
118,000ops/sec
HIGHER IS BETTER
Empirical Conclusion:Direct pointer mapping reduced per-payload latency from 1.84ms to 0.12ms (15.3x speedup) with 92% lower heap allocations.
BM-001RS-001via EX-001

Retrieval Latency & Stability Benchmark

AMD Ryzen / Node v24.20.0 / Single Thread
Baseline Reference:Naive iterative LLM prompt tool loop
KS Lab Candidate:Deterministic State Machine + Inverted Index
Query Latency (p99)
1.15ms
LOWER IS BETTER
State Divergence Rate
0%
LOWER IS BETTER
Throughput
2,380ops/sec
HIGHER IS BETTER
Empirical Conclusion:Candidate achieved 100% state stability with 1.15ms p99 latency versus 820ms baseline.