Pith. sign in

REVIEW 6 cited by

Up or Down? Adaptive Rounding for Post-Training Quantization

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2004.10568 v2 pith:C37Z7BMU submitted 2020-04-22 cs.LG cs.CVstat.ML

classification cs.LGcs.CVstat.ML
keywords lossadaroundpost-trainingquantizationroundingtaskdatafine-tuning
verification ladder T0 review T1 audit T2 compute T3 formal
0 comments
read the original abstract

When quantizing neural networks, assigning each floating-point weight to its nearest fixed-point value is the predominant approach. We find that, perhaps surprisingly, this is not the best we can do. In this paper, we propose AdaRound, a better weight-rounding mechanism for post-training quantization that adapts to the data and the task loss. AdaRound is fast, does not require fine-tuning of the network, and only uses a small amount of unlabelled data. We start by theoretically analyzing the rounding problem for a pre-trained neural network. By approximating the task loss with a Taylor series expansion, the rounding task is posed as a quadratic unconstrained binary optimization problem. We simplify this to a layer-wise local loss and propose to optimize this loss with a soft relaxation. AdaRound not only outperforms rounding-to-nearest by a significant margin but also establishes a new state-of-the-art for post-training quantization on several networks and tasks. Without fine-tuning, we can quantize the weights of Resnet18 and Resnet50 to 4 bits while staying within an accuracy loss of 1%.

Discussion (0). Sign in to comment.

Forward citations

Cited by 6 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. FPTQuant: Function-Preserving Transforms for LLM Quantization

    cs.LG 2025-06 conditional novelty 7.0 of 10

    FPTQuant introduces function-preserving transforms that make transformer activations amenable to static 4-bit quantization with minimal inference overhead.

  2. QuantiBias: Benchmarking Quantization-Induced Bias in LLMs

    cs.CL 2026-07 conditional novelty 6.0 of 10

    Quantization leaves refusal and multiple-choice bias checks flat while open-ended stereotype endorsement remains high (~24–27% under an independent judge), a gap standard safety evaluations miss.

  3. Post-Training Quantization of Generative and Discriminative LSTM Text Classifiers: A Study of Calibration, Class Balance, and Robustness

    cs.LG 2025-07 conditional novelty 6.0 of 10

    Generative LSTM classifiers under post-training quantization are far more sensitive than discriminative ones to calibration data class balance and input noise, especially at 3- to 5-bit widths.

  4. BASE-Q: Bias and Asymmetric Scaling Enhanced Rotational Quantization for Large Language Models

    cs.LG 2025-05 conditional novelty 6.0 of 10

    BASE-Q combines bias correction and asymmetric scaling under fixed rotations to improve 4-bit weight-activation quantization, narrowing the accuracy gap to full precision by up to 50.5% over prior rotation-based methods.

  5. QuantV2X: A Fully Quantized Multi-Agent System for Cooperative Perception

    cs.CV 2025-09 conditional novelty 5.0 of 10

    QuantV2X shows that a fully quantized multi-agent fusion system reduces end-to-end latency by 3.2x and improves system-level mAP30 by 9.5 over a full-precision system on the V2X-Real dataset.

  6. Cross-Layer Error Compensation and Finite-Sample Feature-Statistics Matching for Extreme Low-Bit Quantization of Large Language Models

    cs.NE 2026-07 conditional novelty 4.0 of 10

    A whole-network final-feature and statistics matching objective improves 1.125-bit and 4.125-bit LLM quantization over layer-local and distillation baselines, but its cross-layer mechanism reduces exactly to final-fea...

Pith tools