Skip to main content
Vansh Verma
← Lab

sr-inference

null resultPythonSystems & data structuresJuly 11, 2026

the question

Is stochastic rounding a useful low-precision accumulator for inference?

what came out

No -- it's the worst one. On real Qwen2.5 GEMV, SR is strictly worse than round-to-nearest: a sqrt(n) random walk with no averaging to redeem its unbiasedness. Deterministic Neumaier compensated summation wins 11-15x at lower cost. SR is a training tool, not an inference tool.

method & receipts

  • Result: null result — an honest negative, reported as it came out.
  • Pre-registered — the prediction was written down and committed before the run.

→ read the code and re-run it

github.com/v-code01/sr-inference