When Quantization Breaks Memory: Recurrent-State Write-Back in Low-Precision Temporal Inference

• URL: http://arxiv.org/abs/2609.04490v1
• Date: Thu, 03 Sep 2026 21:23:35 GMT
• Title: When Quantization Breaks Memory: Recurrent-State Write-Back in Low-Precision Temporal Inference
• Abstract summary: In recurrent networks, the quantized state is stored and returned at the next time step, so the rule used to store that state can alter subsequent computations.<n>We introduce recurrent-state write-back to denote this rule and isolate its effect in a compact GRU encoder--decoder for fluorescence lifetime imaging.<n>Our results establish recurrent-state write-back as a key determinant of low-precision recurrent dynamics and identify the state-storage interface as a central design consideration for quantized recurrent inference.
• Score: 0.0
• License: http://creativecommons.org/licenses/by-nc-sa/4.0/
• Abstract: Quantization is widely used to reduce the computational and memory demands of neural-network inference. In recurrent networks, however, the quantized state is stored and returned at the next time step, so the rule used to store that state can alter subsequent computations. Here, we introduce recurrent-state write-back to denote this rule and isolate its effect in a compact GRU encoder--decoder for fluorescence lifetime imaging, a molecular imaging modality used in quantitative biological imaging. A central task is estimating two lifetime parameters, the short-lived component τ1 and the long-lived component τ2, from high-noise time-resolved fluorescence signals. Holding the trained model fixed, replacing continuous state propagation with deterministic 4-bit state storage increases estimation errors for τ1 and τ2 by approximately 70x and 300x, respectively. Failure occurs when repeated small updates remain below the write threshold, leaving the stored state nearly fixed while the network continues to propose change. Error feedback, residual memory, and direction memory carry information from these suppressed updates across time and recover accuracy without retraining. Precision sweeps show that increasing state precision can worsen a fixed recurrent solution, while matched training shows that compatibility with the state interface can be learned. To test whether this behavior extends beyond the GRU, we repeat the post-training intervention in an independently trained LSTM, where coarse write-back reproduces the failure, error feedback restores accuracy, and state-specific interventions reveal greater sensitivity of the cell state than the hidden state. Our results establish recurrent-state write-back as a key determinant of low-precision recurrent dynamics and identify the state-storage interface as a central design consideration for quantized recurrent inference.

Related papers

• ReCal3R: Reliability-Calibrated Learning Rates for Streaming 3D Reconstruction [57.10941470313318]
  Streaming 3D reconstruction relies on a compact recurrent scene state to process long image streams in linear time and bounded memory.<n>We introduce ReCal3R, a reliability-calibrated learning rate method for recurrent 3D reconstruction.
  arXiv  Detail & Related papers  (2026-07-06T17:35:18Z)
• LCPNet: Latent Consistent Proximal Unfolding Network for Infrared Small Target Detection [50.3892176822492]
  Infrared small target detection (IRSTD) aims to identify long distance small targets from complex infrared backgrounds.<n>Deep learning methods have improved IRSTD by learning discriminative image-to-mask mappings.<n>Deep unfolding methods partially address this issue by embedding model-driven iterations into neural networks.<n>We propose Latent Consistent Proximal unfolding network (LCPNet) to address these limitations.
  arXiv  Detail & Related papers  (2026-07-06T02:15:09Z)
• Timestep Rescheduling in Diffusion Inversion [79.30105376216544]
  We show how the selection of the diffusion timestep in the noise scheduler influences inversion fidelity.<n>We propose a simple yet effective nonuniform timestep scheduler that integrates a global rescaling with a local dynamic programming based rescheduling.<n>Our method serves as an off-the-shelf enhancement for existing inversion techniques and requires no extra parameters or computational overhead.
  arXiv  Detail & Related papers  (2026-06-13T16:32:01Z)
• SinkRec: Mitigating Semantic State Sink in Long Sequence Recommendation with Memory-Conditioned Gated Delta Networks [31.41875429254432]
  We propose SinkRec, a hybrid memory-transition looped architecture to mitigate semantic state sink.<n> SinkRec externalizes recurring local patterns into a learnable conditional memory.<n>TDGD uses memory to purify recurrent writing and reading by suppressing memory-covered updates and removing memory-aligned readout responses.
  arXiv  Detail & Related papers  (2026-06-03T09:10:06Z)
• Improving the Performance and Learning Stability of Parallelizable RNNs Designed for Ultra-Low Power Applications [1.226598527858578]
  Bistable Memory Recurrent Unit (BMRU) was introduced to enable hardware-software co-design of ultra-low power RNNs.<n>BMRU performance lags behind parallelizable RNNs on complex sequential tasks.<n>We propose a cumulative update formulation that restores flow while preserving persistent memory, creating skip-connections through time.
  arXiv  Detail & Related papers  (2026-05-12T09:39:33Z)
• Recurrent Quantum Feature Maps for Reservoir Computing [0.6299766708197881]
  We introduce a reservoir based on recurrent quantum feature maps where a fixed quantum circuit is reused to encode both current inputs and a classical feedback signal derived from previous outputs.<n>We evaluate the model on the Mackey-Glass time-series prediction task using our recently introduced CP feature map, and find that it achieves lower mean squared error than standard classical baselines.
  arXiv  Detail & Related papers  (2026-04-03T21:33:10Z)
• Understanding Quantization of Optimizer States in LLM Pre-training: Dynamics of State Staleness and Effectiveness of State Resets [10.325245543844245]
  We study low-precision exponential moving average (EMA) states and show how quantization can cause many nominal updates to round back to the same stored value.<n>We develop a simple predictive model of stalling that estimates one-step stalling probabilities and characterizes how stalling builds up over time.<n>This perspective provides a mechanistic explanation for why-state resets help in low precision.
  arXiv  Detail & Related papers  (2026-03-17T16:14:36Z)
• Never Reset Again: A Mathematical Framework for Continual Inference in Recurrent Neural Networks [2.1838661321884443]
  Recurrent Neural Networks (RNNs) are widely used for sequential processing but face limitations with continual inference due to state saturation.<n>We propose an adaptive loss function that eliminates the need for resets during inference while preserving high accuracy over extended sequences.
  arXiv  Detail & Related papers  (2024-12-20T15:24:28Z)
• Temporal Feature Matters: A Framework for Diffusion Model Quantization [105.3033493564844]
  Diffusion models rely on the time-step for the multi-round denoising.<n>We introduce a novel quantization framework that includes three strategies.<n>This framework preserves most of the temporal information and ensures high-quality end-to-end generation.
  arXiv  Detail & Related papers  (2024-07-28T17:46:15Z)
• Memory Efficient Neural Processes via Constant Memory Attention Block [55.82269384896986]
  Constant Memory Attentive Neural Processes (CMANPs) are an NP variant that only requires constant memory. We show CMANPs achieve state-of-the-art results on popular NP benchmarks while being significantly more memory efficient than prior methods.
  arXiv  Detail & Related papers  (2023-05-23T23:10:19Z)
• EcoTTA: Memory-Efficient Continual Test-time Adaptation via Self-distilled Regularization [71.70414291057332]
  TTA may primarily be conducted on edge devices with limited memory. Long-term adaptation often leads to catastrophic forgetting and error accumulation. We present lightweight meta networks that can adapt the frozen original networks to the target domain.
  arXiv  Detail & Related papers  (2023-03-03T13:05:30Z)
• HiPPO: Recurrent Memory with Optimal Polynomial Projections [93.3537706398653]
  We introduce a general framework (HiPPO) for the online compression of continuous signals and discrete time series by projection onto bases. Given a measure that specifies the importance of each time step in the past, HiPPO produces an optimal solution to a natural online function approximation problem. This formal framework yields a new memory update mechanism (HiPPO-LegS) that scales through time to remember all history, avoiding priors on the timescale.
  arXiv  Detail & Related papers  (2020-08-17T23:39:33Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.