Industrial Internet of Things (IIoT) edge pipelines routinely lose contiguous blocks of multi-sensor data to network dropouts and hardware faults. The standard remedy is a disjointed "impute-then-detect" pipeline whose imputer is trained only for Mean Squared Error (MSE). Such an imputer fills a missing block with the most probable normal pattern: a partially observed anomaly is smoothed back toward normality, and boundary artifacts on normal data raise false alarms. We call this imputation bias and propose the Task-Aware Diffusion Refiner (TADR), a conditional diffusion refiner trained against the loss landscape of a frozen downstream detector. Because back-propagating through the sampling trajectory is intractable, the task-aware terms act on the one-step clean-window estimate that the noise prediction already yields, so training costs no more than plain diffusion training. Under strict Point-to-Point evaluation over five seeds with paired significance tests, TADR raises F1 by 0.338 over the unrefined detector on SMD (p=0.0015), and the anomaly enhancement objective alone contributes 0.463 in ablation. We also characterise when the method helps: the gain tracks the continuous, mutually redundant channels available and the anomaly evidence the outage removes, and it is absent on telemetry datasets dominated by one-hot command channels and on PSM. At ten sampling steps TADR runs in 41.4 ms per window on a single CPU core, sustaining about 24 concurrent 1 Hz streams from a 4.03 MB model. Code and per-seed results are available at https://github.com/int-park422/TADR.
Figures & tables
Fig. 1: Conceptual illustration of imputation bias and pipeline failure, demonstrated using a sampled anomaly region from the normalized Server Machine Dataset (SMD). When an anomaly coincides with a block missing event, standard MSE-based optimization wastes computational resources to synthesize a statistically probable normal pattern, inadvertently erasing the true anomaly. TADR resolves this by directly preserving critical sharp variations for the detector.
Fig. 2: Overall Architecture of the Task-Aware Diffusion Refiner (TADR). The corrupted input X~ is reconstructed by the trainable refiner gϕ . The refined output X^ is then passed through the completely frozen anomaly detector to compute three task-aware objectives, ensuring that gradients flow backward solely to update the parameters ϕ .
Fig. 3: Global P2P F1 across four IIoT datasets at 40% block missing, five seeds with 95% confidence intervals. Datasets are annotated with the number of continuous channels they provide, which is what the benefit of task-aware refinement tracks.
Fig. 4: Refiner comparison under 40% block missing (SMD 1-1). Means over five seeds; error bars are 95% intervals on F1, and the right column the paired difference against the unrefined detector (paired t -test: TADR p=0.0015 , iTransformer p=0.0010 , CSDI p=0.018 , TIMBA p=0.028 ).
Masked fraction
Points
Raw
CSDI
TADR
25–50%
2,643
0.344
0.735
0.778
50–75%
51
0.205
0.665
0.830
75–100%
0
—
—
—
Normal (false alarms)
25,785
0.027
0.056
0.037
TABLE I: Detection rate by masked fraction, the share of covering windows in which an anomalous point was dropped (SMD 1-1, 40% block missing, five seeds). The last row is the false-alarm rate on normal points.
Fig. 5: Detector generality on pooled SMD. Bars are means over five seeds, error bars 95% intervals on the refined score, and the annotation above each pair the paired difference. Gains are significant for KAN-AD and TimesNet ( p≤0.002 ), absent for Anomaly Transformer.
Channels retained
38
24
12
6
3
1
Δ F1 vs. Raw
+0.356
+0.293
+0.217
+0.168
+0.050
+0.041
p
< 0.0001
0.004
0.015
0.049
0.51
0.29
Test stride
1
5
25
50
100
Δ F1 vs. Raw
+0.352
+0.372
+0.380
+0.339
+0.300
p
< 0.0001
< 0.0001
< 0.0001
< 0.0001
0.0004
TABLE II: Benefit against the number of channels retained (SMD 1-1, 40% block missing), and against the overlap between successive evaluation windows. Δ is the paired difference over five seeds.
Loss Components
P2P F1
Precision
Recall
Δ
p
Recon (MSE Only)
0.306±0.069
0.525
0.313
—
—
Recon + Feature
0.542±0.091
0.561
0.540
+0.236
0.012
Recon + Feature + Enhance (Full)
0.769±0.049
0.762
0.787
+0.463
0.0002
Full objective, single-pass refiner
0.469±0.065
0.708
0.359
+0.163
0.010
TABLE III: Ablation on loss components (smoothed P2P F1, 40% block missing). Δ is the paired difference against the MSE-only refiner over five seeds.
Fig. 6: P2P F1 over the (λfeat,λenhance) grid, mean of five seeds. The setting used throughout the paper is outlined.
Configuration
ms / batch
ms / window
Raw (No Refiner)
0.39
0.002
TIMBA
0.85
0.003
iTransformer
1.67
0.007
CSDI ( T=50 )
241.8
0.944
Ours (TADR), T=50
240.5
0.939
Ours (TADR), T=10
44.2
0.173
TABLE IV: Inference overhead. Batch figures use batch size 256 on one RTX 3090 with CUDA synchronisation, measured with the GPU otherwise idle (means of five runs; the T=10 row is from the sampling-step profile). Streaming rows use batch size 1 on a single CPU thread of a 2.1 GHz Xeon E5-2620 v4. The refiner occupies 4.03 MB.
T
CSDI F1
TADR F1
TADR sat. (%)
TADR obs. MSE
50
0.628
0.710
0.0
1.86
25
0.564
0.714
0.0
1.64
10
0.426
0.747
0.0
1.96
5
0.452
0.710
0.0
6.16
2
0.435
0.695
0.9
63.0
TABLE V: Detection quality and reconstruction fidelity against the sampling budget (SMD 1-1, 40% block missing). Saturation is the fraction of output pinned to the clipping bound; observed MSE is over never-missing samples.
Diffusion models have achieved outstanding performance in unsupervised industrial anomaly detection (uIAD) by learning a manifold of normal data under the common assumption that off-manifold anomalies are harder to generate, resulting in larger reconstruction errors in data space or lower probability densities in the tractable latent space. However, their iterative denoising and noising nature leads to slow inference. In this paper, we propose OSD-IRF, a novel one-step diffusion with inverse residual fields, to address this limitation for uIAD task. We first train a deep diffusion probabilistic model (DDPM) on normal data without any conditioning. Then, for a test sample, we predict its inverse residual fields (IRF) based on the noise estimated by the well-trained parametric noise function of the DDPM. Finally, uIAD is performed by evaluating the probability density of the IRF under a Gaussian distribution and comparing it with a threshold. Our key observation is that anomalies become distinguishable in this IRF space, a finding that has seldom been reported in prior works. Moreover, OSD-IRF requires only single step diffusion for uIAD, thanks to the property that IRF holds for any neighboring time step in the denoising process. Extensive experiments on three widely used uIAD benchmarks show that our model achieves SOTA or competitive performance across six metrics, along with roughly a 2X inference speedup without distillation.
Boan Zhang, Wen Li, Guanhua Yu +3
Department of Computer Science and Technology Xidian University Shaanxi, China 710126 · Department of Electronic Engineering Xidian University Shaanxi, China 710126
Continual anomaly detection (CAD) addresses the need for industrial inspection systems to adapt to evolving production conditions, yet existing methods share three critical gaps: unrealistic evaluation, no systematic comparison, and no consideration of edge deployment constraints. We introduce a unified benchmark combining discrete-task evaluation on structural and logical anomalies, a novel continuous drift protocol, the first head-to-head comparison of all published CAD methods, and computational efficiency profiling on edge hardware. Our results reveal that existing CAD methods do not consistently outperform traditional approaches with simple experience replay. Thus motivated, we propose DINOSaur, a training-free method combining a frozen DINOv3 backbone with spatially-indexed coreset memory and neighborhood-restricted anomaly scoring. DINOSaur achieves zero forgetting by construction, outperforms all evaluated methods across all five protocols, and runs at sub-100,ms inference on an NVIDIA Jetson Orin Nano, with on-device adaptation to new tasks in under 30 seconds.
Latent defect screening is challenged by extremely low failure rates, high-dimensional test data, and absence of labeled anomalies. We propose the first unsupervised anomaly detection framework incorporating a Diffusion Transformer. Raw test measurements are first compressed by an autoencoder, then reshaped into a structured token sequence enriched with sinusoidal and per-device wafer-position embeddings. Anomaly scores are derived from the noise-prediction error over mid-range diffusion timesteps, enabling fast wafer-scale screening without any labeled defects or manual feature engineering. Our approach achieves state-of-the-art performance on industrial 16nm IC test data under extreme class imbalance, offering interpretable failure localization through latent-space reconstruction residuals.
Yuxuan Yin, Chen He, Todd Jacobs +4
Department of Electrical and Computer Engineering, University of California Santa Barbara, CA, USA · Automotive Processing, NXP Semiconductors, TX, USA