TheMachine Press

A daily newspaper for the age of artificial intelligence.

Morning editionPermanent story

chips infrastructure

Four-Bit Memory Turned Small Errors Into a Three-Hundredfold Miss

A recurrent network kept proposing updates too small to survive storage, freezing its state until residual and direction memory restored the lost information.

Published Updated Story ID: mp-2026-09-07-005
Read the complete editionStory JSON

Summary

A recurrent network kept proposing updates too small to survive storage, freezing its state until residual and direction memory restored the lost information.

Holding a trained GRU fixed while replacing continuous state propagation with deterministic four-bit storage increased two fluorescence-lifetime estimation errors by about 70 and 300 times. The authors trace the failure to small repeated updates falling below the write threshold. Error feedback, residual memory and direction memory recovered accuracy without retraining, and an independently trained LSTM reproduced the pattern. The measurements come from the tested imaging models and do not establish a universal penalty for four-bit inference.

Why it matters

A recurrent network kept proposing updates too small to survive storage, freezing its state until residual and direction memory restored the lost information.

Limits and context

  • The measurements come from the tested imaging models and do not establish a universal penalty for four-bit inference.

Key claims

  1. A recurrent network kept proposing updates too small to survive storage, freezing its state until residual and direction memory restored the lost information.

    Qualification: The measurements come from the tested imaging models and do not establish a universal penalty for four-bit inference.

    Evidence: source-2026-09-07-005

Sources

  1. arXiv preprint 2609.04490arXiv · primary research

Corrections

No corrections have been recorded for this story.