--- title: 'GLOD: Multi-Domain Global-Local Approaches' url: https://www.emergentmind.com/topics/glod type: topic --- # GLOD: Multi-Domain Global-Local Approaches Searching arXiv for recent and canonical uses of “GLOD” across domains. Searching for graph-level OOD/outlier detection usages of GLOD. Searching for computer vision usages of GLOD/GLoD. In current arXiv usage, **GLOD** is not a single standardized term. It most prominently denotes **Graph-Level Outlier Detection** and **Graph-Level Out-of-Distribution detection** in graph machine learning, but it is also used for the **Global-Local Object Detector** in satellite imagery, for **GRAND with log-likelihood outputs** in joint detection and decoding, and, in the capitalization variant **GLoD**, for **Global-Local Diffusion** in image generation and the **Global-Local Dataset** for text-rich image translation [2012.12931] [2406.15523] [2507.11040] [2207.10836] [2404.15447] [2602.21956]. ## 1. Terminological scope The literature uses the label “GLOD” for several technically unrelated constructs. In graph mining, it refers either to graph-level outlier detection or graph-level OOD detection. In remote sensing, it names a transformer-first detector for high-resolution satellite imagery. In coding theory, it denotes a soft-output GRAND variant. In generative modeling and multimodal translation, the capitalization variant “GLoD” denotes a diffusion framework and a dataset, respectively [2012.12931] [2406.15523] [2507.11040] [2207.10836] [2404.15447] [2602.21956]. | Usage | Domain | Representative source | |---|---|---| | Graph-Level Outlier Detection / Graph-Level OOD detection | Graph machine learning | [2012.12931], [2406.15523] | | Global-Local Object Detector | Remote sensing object detection | [2507.11040] | | GRAND with log-likelihood outputs | Channel decoding | [2207.10836] | | Global-Local Diffusion | Text-to-image generation | [2404.15447] | | Global-Local Dataset | Text Image Machine Translation | [2602.21956] | A common source of confusion is the proximity between **GLOD** and **GOLD**. They are distinct: GOLD is the NASA **Global-scale Observations of the Limb and Disk** ultraviolet imaging spectrograph, whereas GLOD/GLoD in the works discussed here denotes other concepts entirely [1912.02271]. ## 2. GLOD in graph machine learning In the 2020 graph-mining formulation, **Graph-Level Outlier Detection (GLOD)** is defined over a database of graphs $\mathcal{G}=\{G_1,\dots,G_N\}$, where each graph is $G=(V,E,X)$ with nodes $V$, edges $E$, and node features $X\in\mathbb{R}^{n\times d}$. The task is to find graphs whose distributional mechanism deviates from that generating the majority of graphs. This differs from node- or edge-level anomaly detection because the entire graph is the scored unit [2012.12931]. A common evaluation protocol repurposes binary graph classification datasets. If the original classes are $C_0$ and $C_1$, one class is selected as inlier, the other is downsampled by a fraction $\alpha\in(0,1]$, and the combined set is evaluated as an outlier-detection benchmark. The study emphasizes that a fair evaluation repeats this in both directions—$C_0$ as inlier and $C_1$ downsampled, then vice versa—and averages over multiple random downsamplings. It also distinguishes **ROC-AUC**, defined as $P(s(G^+)>s(G^-))$, from **PR-AUC**, which is more sensitive under heavy imbalance [2012.12931]. The later UB-GOLD benchmark broadens the term toward **Graph-Level Out-of-Distribution detection**, with unlabeled ID training graphs $D_{\text{train}}=\{G_1,\dots,G_n\}$ drawn from $P^{in}$ and test graphs drawn from a mixture of $P^{in}$ and $P^{out}$. The detector outputs an OOD score $s(G)=f(G)$, and the graph-level decision rule is $$ y_{OOD}(G)=\mathbb{1}[s(G)>\tau]. $$ UB-GOLD treats both GLAD and GLOD as instances of generalized graph-level OOD detection under a single unsupervised objective [2406.15523]. ## 3. Evaluation pathologies, benchmarks, and methodological consequences A central empirical result in graph-level outlier detection is the **performance flip** observed when binary graph classification datasets are repurposed for GLOD. Across many datasets and detectors, ROC-AUC can be high when one class is downsampled yet become very low—sometimes worse than random—when the roles of the two classes are swapped. Representative values at $\alpha=0.1$ include, on **DD**, WL+LOF: $0.186$ versus $0.815$, PK+LOF: $0.194$ versus $0.824$, and OCGIN: $0.327$ versus $0.720$; on **NCI1**, the direction can reverse, as with WL+LOF: $0.730$ versus $0.349$ [2012.12931]. The same study reports an approximate **AUC complementarity**: $$ AUC(1\to \text{outlier}) + AUC(0\to \text{outlier}) = 1 - P[s(X)=s(Y)], $$ so the two AUCs often sum to approximately $1$ when ties are negligible. It further attributes the flip to two embedding-space effects: **density disparity**, measured with neighborhood diagnostics such as NN-Radius, and **overlapping support**, quantified by NN-Disagreement%. For propagation-based methods such as the Weisfeiler–Lehman subtree kernel and the Propagation Kernel, increasing propagation depth $L$ amplifies the gap between the two downsampling variants because sparsification and diagonal dominance increase with $L$ [2012.12931]. UB-GOLD reframes these issues at benchmark scale. It covers **35 datasets** across **four practical scenarios**: intrinsic anomaly, class-based anomaly, inter-dataset shift, and intra-dataset shift, and compares **18 representative GLAD/GLOD methods**. It evaluates detectors with **AUROC**, **AUPRC**, and **FPR@95**, and conducts multi-dimensional analyses of effectiveness, OOD sensitivity spectrum, robustness, and efficiency. Aggregate AUROC results reported in the benchmark place **SIGNET** at **75.86** average AUROC with average rank **3.34**, followed by **OCGTL** at **73.14**, while end-to-end methods consistently outperform two-step baselines on average [2406.15523]. Several practical consequences follow directly from these studies. The 2020 work recommends evaluating both downsampling choices, using mixtures of “X vs. Non-X” and “X vs. Y” datasets, reporting PR-AUC in addition to ROC-AUC, and tempering propagation-induced sparsification. UB-GOLD similarly emphasizes multi-metric evaluation, unsupervised threshold calibration, robustness to contaminated training data, and the fact that **near-OOD** detection remains substantially harder than **far-OOD** detection [2012.12931] [2406.15523]. ## 4. GLOD as the Global-Local Object Detector In remote sensing, **GLOD** denotes the **Global-Local Object Detector**, a transformer-first architecture for object detection in high-resolution satellite imagery. It replaces conventional CNN backbones with a **Swin Transformer** encoder, uses a convolutional neck built from **UpConvMixer (UCM) blocks** and **Fusion Blocks**, and ends with a **CenterNet-inspired anchor-free head**. The design targets tiny objects, severe scale variation and density, and computational constraints in overhead imagery [2507.11040]. The end-to-end pipeline is defined for a fixed input of **3072×3072**. The Swin backbone produces hierarchical stages **S1–S4** by windowed self-attention and shifted windows. UCM blocks then upsample and refine features using asymmetric fusion, separable convolutions, CBAM attention, Highway gating, and PixelShuffle. The asymmetric fusion module is given as $$ AF_X=\operatorname{ReLU}\!\left(\sum_{k\in\{1\times3,\,3\times3,\,3\times1\}} BN(k*X)\right), $$ and the UCM recurrence with $N=3$ repeats includes $$ X_3^{(i)}=g\cdot X_2^{(i)} + (1-g)\cdot X_3^{(i-1)},\qquad Y_{UCM}=\operatorname{PixelShuffle}(X_3^{(i)}). $$ Fusion Blocks combine adjacent UCM outputs by channel alignment, bilinear upsampling of the lower-resolution stream, and additive fusion followed by GELU [2507.11040]. The prediction head outputs center heatmaps, sub-pixel offsets $(\Delta x,\Delta y)$, and sizes $(w,h)$. The total loss is $$ \mathcal{L}_{total}=\lambda_{cls}\mathcal{L}_{cls}+\lambda_{off}\mathcal{L}_{off}+\lambda_{size}\mathcal{L}_{size}, $$ with $\lambda_{cls}=\lambda_{off}=\lambda_{size}=1$. Training on **xView** uses **AdamW** with $\alpha=5\times10^{-5}$, $\beta_1=0.9$, $\beta_2=0.999$, $\epsilon=10^{-8}$, cosine warm restarts, **42k training steps**, per-device batch size **3**, accumulated batch size **24**, and gradient checkpointing. Images are resized to **3072×3072**, normalized with ImageNet-1K mean and standard deviation, and trained from scratch rather than with backbone pretraining [2507.11040]. Reported xView performance is **32.95 mAP50** and **28.87 mAP75**. This exceeds the best CNN baseline listed, **FPN-50 RFL** at **28.32 mAP50**, by **4.63** points, and the best transformer baseline listed, **DQ-DETR** at **29.56 mAP50**, by **3.39** points. Ablation also attributes a substantial gain to Fusion Blocks: with FB, **32.95/28.87** mAP50/mAP75; without FB, **30.13/24.72** [2507.11040]. ## 5. GLoD as Global-Local Diffusion In image generation, the capitalization variant **GLoD** denotes **Global-Local Diffusion**, a training-free framework for text-to-image generation that composes **global prompts** describing scene-level interactions with **local prompts** specifying object-specific attributes. The method is designed for the case where diffusion models misattribute attributes across multiple entities or ignore them when a single prompt must encode both global interactions and local details [2404.15447]. GLoD uses a pre-trained diffusion model and performs multiple forward passes at each timestep to obtain unconditional, global-conditioned, and local-conditioned predicted noises. Its global guidance is $$ g_g=\epsilon_\theta(_t,t|c_g)-\epsilon_\theta(_t,t), $$ while its local guidance for object $j$ within mask $M_j$ is $$ g_l=M_j\big(\epsilon_\theta(_t,t|c_j)-\epsilon_\theta(_t,t|c_b)\big). $$ Sampling then follows $$ _{t-1}\sim \mathcal{N}\big(_t-(\epsilon+g_g+g_l),\sigma_t^2 I\big). $$ The crucial design choice is that local guidance is defined **relative to the base global condition** inside the object mask, rather than as an independent prompt added uniformly across the image [2404.15447]. The framework is implemented with **Stable Diffusion v2.1**, latent space $\mathbb{R}^{64\times64\times4}$, Euler Discrete Scheduler, and training-free layout control via backward guidance using cross-attention maps and masks. Runtime scales linearly with the number of prompts because each step requires **1 unconditional + (#global) + (#local)** U-Net evaluations. Bounding-box masks are central to reducing attribute leakage, and the paper defines an **infection score** $S_i$ to quantify undesirable cross-effects between objects [2404.15447]. On a **2,500-sample** controllability-oriented test set, GLoD improves the combined global-local score **$S_{gl}$** and reduces infection relative to training-free layout control. Reported values include training-free layout control at approximately **$S_g\approx24.4$**, **$S_{ls}\approx21.2$**, **$S_{lo}\approx20.5$**, **$S_{gl}\approx22.6$**, **$S_i\approx15.6$**, whereas GLoD reports **$S_g\approx24.6$**, **$S_{ls}\approx23.3$**, **$S_{lo}\approx20.9$**, **$S_{gl}\approx23.3$**, **$S_i\approx14.5$** [2404.15447]. ## 6. GLOD as GRAND with log-likelihood outputs In coding theory, **GLOD** denotes **GRAND with log-likelihood outputs**, introduced in a soft-input, soft-output joint detection and GRAND framework. GRAND itself performs maximum-likelihood decoding by guessing additive noise sequences in decreasing order of likelihood and returning the first noise-recovered word that lies in the codebook. GLOD extends this procedure to produce bitwise **log-likelihood ratios (LLRs)** using Euclidean-distance metrics computed for candidate noise-recovered words [2207.10836]. For bit $b_k$, the target quantity is $$ \mathrm{LLR}(b_k)=\log\frac{\Pr(b_k=1\mid y)}{\Pr(b_k=0\mid y)}. $$ The method approximates the required log-sum-exp using the small set of candidates found through GRAND’s guesswork. In its Max-Log form, it tracks the current best distance $d^{ML}$ and a per-bit counter-ML distance $d^{cML}_{i,j}$, yielding $$ \bar\lambda^{GRAND}_{i,j}\approx \begin{cases} \dfrac{1}{\sigma^2}\big(d^{cML}_{i,j}-d^{ML}\big), & \hat c^{GRAND}_{i,j}=1,\\[4pt] \dfrac{1}{\sigma^2}\big(d^{ML}-d^{cML}_{i,j}\big), & \hat c^{GRAND}_{i,j}=0. \end{cases} $$ The paper emphasizes that the entropy of noise is typically smaller than the entropy of information bits, so a relatively small number of guesses often suffices to obtain reliable LLRs for critical bits [2207.10836]. GLOD is also embedded in an iterative **turbo-GRAND** loop. Extrinsic LLRs produced in one iteration are reused as a priori information to reorder noise guesses in the next iteration, improving both detection and decoding. Simulations for **BCH [127,113]** in **AWGN** and **Rayleigh fading** show that a few turbo-GRAND iterations match the performance of ML-detection-based soft-GRAND while average complexity grows linearly rather than exponentially with the number of symbols. The reported gains include **more than 6 dB SNR advantage at BLER $10^{-2}$** in Rayleigh fading over schemes lacking quality soft information, with two iterations capturing most of the improvement [2207.10836]. ## 7. GLoD as the Global-Local Dataset In multimodal translation, **GLoD** stands for **Global-Local Dataset**, introduced as the training substrate for **GLoTran** in high-resolution, text-rich **Text Image Machine Translation (TIMT)**. A GLoD sample couples a low-resolution **global image** $I_g$ with a set of **local slices** $\mathcal{I}=\{I_1,\dots,I_{N_s}\}$ obtained by detecting, grouping, and cropping text regions from the original image, along with source-language text, target-language translation, bounding boxes, reading order, and grouping metadata [2602.21956]. The dataset is reported as **517,354 global-local image–text pairs across 517,354 images**, spanning **more than 40 real-world scenarios** and **5 languages**. Scenarios explicitly include documents, posters, menus, road signs, receipts, photos, and other text-rich settings with irregular layouts, small and scattered text, handwriting, decorative fonts, and non-text distractors. The curation pipeline comprises scene conceptualization, data collection and pre-filtering, text region detection and grouping, global-local translation, and quality control. An intermediate stage initially yields approximately **2M global-local pairs** before rigorous filtering [2602.21956]. Its formal structure is tied to the global-local translation model. Starting from an image $I$, a detector produces bounding boxes $B=\{b_i\}_{i=1}^{N_b}$, each local crop is $\tilde I_i=\operatorname{Crop}(I,b_i)$, an ordering function yields $\tilde I^{ord}=ord(\{\tilde I_i\})$, and a merge operator gives $\mathcal{I}=M(\tilde I^{ord})=\{I_i\}_{i=1}^{N_s}$ with $N_s\ll N_b$. GLoTran then consumes the ordered local slices together with the global image and special tokens $\langle image_g\rangle$ and $\langle image_l\rangle$, using a replay window of previous translations to maintain consistency. The training objective is standard cross-entropy over target tokens: $$ \theta^*=\arg\min_\theta \frac{1}{|D|}\sum_{(I,Y)\in D}\left[-\sum_{i=1}^{N_s} Y_i \log p(\hat Y_i\mid I_g,I_i,P_{