Why Sub-4-Bit Vision CNNs Collapse and Where Mixed-Precision Recovery Stops
Organizations: GE Appliances, a Haier company
Abstract
Deep vision models operating in resource-constrained environments depend heavily on low-bit quantization, yet compressing weights below 4 bits frequently induces catastrophic accuracy collapse. While mixed-precision recovery strategies aim to rescue accuracy by selectively retaining high-precision parameters, the fundamental mechanics of where and why these models collapse remain poorly understood. In this study, we systematically analyze the anatomy of quantization collapse across standard Vision CNNs (including ResNet and EfficientNet backbones) evaluated on the full 50,000-image ImageNet-1k validation set. First, we map the exact progression of degradation, demonstrating that collapse is highly non-linear and concentrated in late-stage pointwise projections and depthwise layers. Second, we introduce a structured, multi-step recovery waterfall that evaluates selective integer precision (INT4/INT8) and selective floating-point scales. We show that keeping less than 5% of critical layer activations in 8-bit precision recovers up to 92% of the accuracy lost during aggressive sub-4-bit compression. Finally, we bridge the hardware reality gap by evaluating deployment performance on standard ARM and x86 CPU architectures. We show that while low-bit quantization reduces memory storage requirements, standard runtime software bit-unpacking introduces a 1.48x to 21.5x execution slowdown depending on batch size, highlighting a critical trade-off between model footprint and actual edge throughput. Our findings provide a practical blueprint for navigating the limits of sub-4-bit deployment without sacrificing hardware performance.
Figures & tables
| Experiment Group | Quantized Scope / Precision Profile | Mean Bits | Top-1 (%) | vs. FP32 (pp) |
|---|---|---|---|---|
| Reference | Unquantized FP32 Baseline (B0 e86) | 32.00 | 72.87 | |
| Single-Mask | Stage 1 Only at W3 (rest FP32) | 31.99 | 25.25 | |
| Perturbations | Early Depthwise Only (Stem + Stages 1–2) at W3 | 31.97 | 31.20 | |
| (50k Verified) | Early Depthwise Only (Stem + Stages 1–2) at W4 | 31.97 | 40.66 | |
| Stage 1 Only at W4 (rest FP32) | 31.99 | 42.96 | ||
| Stage 2 Only at W3 (rest FP32) | 31.90 | 48.88 |
| Precision Profile | Bits | Size | MaxAbs+Src | MSE+Src | MaxAbs+BN | MSE+BN | +1Ep QAT+BN |
|---|---|---|---|---|---|---|---|
| FP32 Reference (B0 e86) | 32.00 | 20.43 MB | 72.86% | — | 70.52% | — | 71.53% (ctrl) |
| Uniform W8 Reference | 8.00 | 5.29 MB | 72.81% | — | 70.50% | — | — |
| Heuristic Tiered W8/W4/W3 | 3.55 | 2.35 MB | 51.15% | 60.15% | 60.15% | 64.09% ( ) | — |
| Adaptive W8/W4/W3 | 3.56 | 2.35 MB | 64.61% | 66.22% | 66.40% | 67.94% ( ) | 68.98% ( ) |
| Adaptive W8/W4/W3 | 3.49 | 2.30 MB | 63.30% | 64.86% | 65.81% | 67.34% ( ) | — |
| Adaptive W8/W4/W3 | 3.40 | 2.24 MB | 61.51% | 63.37% | 64.57% | 66.35% ( ) | — |
| Runtime Path | Precision Format | Size | Size | Top-1 | CPU p50 / p95 (ms) | Execution Mechanism |
|---|---|---|---|---|---|---|
| LiteRT (TFLite) | Float32 Baseline | 20.43 MB | 72.86% | 9.81 / 10.98 | Native FP32 SIMD kernels | |
| LiteRT (TFLite) | Full-Integer INT8 | 6.00 MB | 72.81% | 6.65 / 7.22 ( ) | Native INT8 SIMD (82 ops) | |
| LiteRT (TFLite) | Mixed 4-Bit (WI4) | 3.59 MB | 63.27% | 15.43 / 16.43 ( ) | 66 INT4 + 16 FP32 DW; unpack | |
| Custom C++/TF | Packed Linear INT3 (Unif.) | 2.47 MB | 0.30% | 376.8 / 386.6 ( ) | Bitstream unpack to FP32 |
Low-Bit Quantization
Low-bit quantization reduces neural-network precision to save memory and computation while seeking to preserve task performance.
See all 254 papers