Falcon-X: X-ray Benchmark & Time Series Model

• Falcon-X is a dual-purpose label used for an X-ray benchmark in compositional threat reasoning and a time series foundation model in heterogeneous forecasting.
• In the X-ray domain, Falcon-X employs structured supervision, dense grounding, and counterfactual extensions to support uncertainty-aware threat inference.
• For forecasting, Falcon-X leverages unified prototype diff-attention and latent entity attention to enable efficient multivariate modeling and reduced computational complexity.

Falcon-X is not a single universally defined system in the arXiv record. The name is used for at least two distinct research artifacts: a benchmark for compositional threat reasoning in dual-energy X-ray baggage screening, introduced alongside the Falcon model for structured safety reasoning, and a time series foundation model for heterogeneous multivariate forecasting that operates in a shared latent prototype space (Michael et al., 24 Jun 2026, Liu et al., 26 May 2026). The term therefore denotes domain-specific constructs rather than a unitary architecture, and its meaning depends on the surrounding research context.

1. Nomenclature and scope

In current arXiv usage, the name has two principal referents. One is Falcon-X as a benchmark in safety-critical X-ray vision-language research; the other is Falcon-X as a time series foundation model in forecasting research. The shared label does not imply a common lineage, objective, or implementation substrate (Michael et al., 24 Jun 2026, Liu et al., 26 May 2026).

Falcon-X usage	Domain	Role
Falcon-X	X-ray baggage screening	Benchmark for compositional threat reasoning
Falcon-X	Time series forecasting	Foundation model for heterogeneous multivariate modeling

This bifurcation is central to interpreting the literature. In the X-ray setting, Falcon-X is an evaluation suite and supervision framework paired with a separate model named Falcon. In the forecasting setting, Falcon-X is itself the model. A plausible implication is that references to “Falcon-X” without domain qualifiers are intrinsically ambiguous.

2. Falcon-X in X-ray compositional threat reasoning

In “Falcon: Functional Assembly and Language for Compositional Reasoning in X-ray”, Falcon-X is a dual-energy X-ray benchmark for evaluating whether a multimodal system can infer threat from the functional assembly potential of spatially dispersed components such as a battery, detonator, and main charge, rather than merely recognize isolated objects (Michael et al., 24 Jun 2026). The underlying problem is formalized as compositional threat reasoning, where risk is a relational property over grounded regions.

The benchmark instantiates a component taxonomy

$$\mathcal{C} = \{\text{battery}, \text{detonator}, \text{main charge}\},$$

and defines a binary component-presence vector

$$\mathbf{y} \in \{0,1\}^{|\mathcal{C}|}, \qquad y_c = 1 \iff \text{at least one instance of component } c \text{ is present}.$$

It also defines a type-level compatibility template

$$\mathbf{L} \in [0,1]^{|\mathcal{C}| \times |\mathcal{C}|},$$

where $L_{uv}$ encodes the functional compatibility strength between component types $u$ and $v$. The pair $(\mathbf{y}, \mathbf{L})$ is the benchmark’s basic structured safety state.

Falcon-X includes a minimal completeness notion,

$$\text{complete}(I) = \mathbb{1}\{\mathbf{y} = \mathbf{1}\},$$

but the paper explicitly distinguishes completeness from risk. Scene-level risk is modeled as a continuous variable

$r \in [0,1],$

and is not trivially derived from the visible presence vector. Even scenes missing one component may receive a high risk score because a part may be occluded, concealed, or ambiguously visible. This makes Falcon-X a benchmark for uncertainty-aware compositional safety reasoning, not deterministic set completion.

The benchmark is explicitly positioned against object-centric evaluation. Standard image captioning, object detection, simple VQA, and referring segmentation can all succeed while still failing the targeted safety problem: inferring whether dispersed grounded parts are jointly sufficient and functionally compatible for a plausible assembly. Falcon-X is therefore presented as the first X-ray benchmark that jointly supports dense grounding, multimodal understanding, and structured functional threat reasoning over dismantled components.

3. Benchmark design, supervision, and evaluation in the X-ray setting

Falcon-X combines dense instance grounding with structured functional labels and multimodal task supervision (Michael et al., 24 Jun 2026). Each image includes bounding boxes and pixel masks, together with component presence, functional completeness, pairwise functional compatibility links, and a scene-level risk score. The associated task suite spans captioning, VQA, referring segmentation, panoptic tasks, functional grounding, missing-component reasoning, and risk prediction.

The dataset statistics are reported with a small unresolved discrepancy. The main paper says Falcon-X contains about 7,000 real dual-energy baggage scans, the appendix states that 7,000 base X-ray scans were collected, and the benchmark table reports 6,911 samples. The counterfactual extension expands the corpus to approximately 50,000 images. Stage 1 uses 16,580 annotated component instances, while Stages 2 and 3 use 442,287 training instructions. Data collection used an ANER K8065 dual-energy X-ray scanner with settings 100–160 kV, 0.4–1.2 mA. The documented split is 80/20 train/test, with all counterfactuals from a base image kept in the same partition; no separate validation split is specified.

A notable feature is the counterfactual extension. Components are selectively removed using mask-guided inpainting, producing controlled variants spanning single components, partial assemblies, and complete configurations. This supports evaluation of missing-component identification, partial-assembly reasoning, functional completeness, and risk recalibration. Risk labels are generated from component presence, type-level compatibility, and visual uncertainty, then expert-verified using a rubric in which $[0.0,0.3)$ is low risk, $$\mathbf{y} \in \{0,1\}^{|\mathcal{C}|}, \qquad y_c = 1 \iff \text{at least one instance of component } c \text{ is present}.$$0 is medium or ambiguous risk, and $$\mathbf{y} \in \{0,1\}^{|\mathcal{C}|}, \qquad y_c = 1 \iff \text{at least one instance of component } c \text{ is present}.$$1 is high functional risk; 72.3% of generated labels were accepted without change and 27.7% were corrected by experts.

Falcon-X organizes evaluation into three layers. Layer I covers grounded perception under X-ray superposition, including scene captions, panoptic segmentation, referring segmentation, and VQA. Layer II covers compositional functional reasoning, including Missing Component Identification, Functional Completeness, and Referring Functional Grounding. Layer III covers relationally consistent safety inference, including scene-level risk prediction, functional link estimation, and logically coherent component-set analysis. Metrics vary by task type: BLEU, METEOR, ROUGE-L, and CIDEr for language; cIoU and mIoU for grounding and segmentation; exact-match accuracy, MAE, and binary accuracy for VQA; Accuracy and F1 for classification-style reasoning; and MAE or RMSE for functional completeness, scene risk, and link-risk regression.

Empirically, the benchmark exposes a clear gap between perceptual adaptation and compositional reasoning. After fine-tuning, many baselines become strong on captioning, VQA, and presence recognition, but remain weak on missing-part reasoning, pairwise compatibility, scene-level risk, and functionally defined grounding. On Referring Functional Grounding, Falcon exceeds the next best model by +30.38 cIoU and +40.03 mIoU. The structured metrics reported for Falcon are CPC Acc = 98.1, FC MAE = 0.017, FC RMSE = 0.09, SRL MAE = 0.02, and CLR MAE = 0.005. The benchmark’s diagnostic value lies precisely in showing that high object presence accuracy does not imply coherent link estimation or calibrated threat inference.

4. Falcon-X as a time series foundation model

In “Falcon-X: A Time Series Foundation Model for Heterogeneous Multivariate Modeling”, Falcon-X is an encoder-only Transformer TSFM for heterogeneous multivariate forecasting (Liu et al., 26 May 2026). Its core claim is that prior TSFMs remain either effectively univariate or perform cross-variate interaction too directly in the raw variate space, where heterogeneous channels from different datasets are not semantically aligned and where standard non-negative attention cannot explicitly represent antagonistic dependencies.

The formal setup is

$$\mathbf{y} \in \{0,1\}^{|\mathcal{C}|}, \qquad y_c = 1 \iff \text{at least one instance of component } c \text{ is present}.$$2

where entity $$\mathbf{y} \in \{0,1\}^{|\mathcal{C}|}, \qquad y_c = 1 \iff \text{at least one instance of component } c \text{ is present}.$$3 is a multivariate time series with $$\mathbf{y} \in \{0,1\}^{|\mathcal{C}|}, \qquad y_c = 1 \iff \text{at least one instance of component } c \text{ is present}.$$4 variates and look-back window $$\mathbf{y} \in \{0,1\}^{|\mathcal{C}|}, \qquad y_c = 1 \iff \text{at least one instance of component } c \text{ is present}.$$5, and the total aggregated variate count is

$$\mathbf{y} \in \{0,1\}^{|\mathcal{C}|}, \qquad y_c = 1 \iff \text{at least one instance of component } c \text{ is present}.$$6

Falcon-X learns a dimension-agnostic mapping

$$\mathbf{y} \in \{0,1\}^{|\mathcal{C}|}, \qquad y_c = 1 \iff \text{at least one instance of component } c \text{ is present}.$$7

Its defining move is to decouple physical variates from the interaction space. Rather than mixing raw channels directly, Falcon-X projects each variate into a small shared set of latent prototypes. This is intended to provide a semantic coordinate system in which variates with different physical meanings can nonetheless align to reusable structural patterns. The paper argues that this yields both semantic alignment and zero-shot structural transfer.

The largest model has about 591M parameters. It first builds per-variate temporal representations, then applies Unified Prototype Diff-Attention, which explicitly computes both positive/synergistic and negative/antagonistic semantic affinities. Cross-variate interaction is then performed in the shared prototype space by Latent Entity Attention, after which a Variate Reassembly Router reconstructs variate-specific representations for forecasting. This produces a multivariate TSFM intended to be more expressive and more scalable than raw-space multivariate attention, with prototype-space interaction reducing complexity from $$\mathbf{y} \in \{0,1\}^{|\mathcal{C}|}, \qquad y_c = 1 \iff \text{at least one instance of component } c \text{ is present}.$$8 to approximately $$\mathbf{y} \in \{0,1\}^{|\mathcal{C}|}, \qquad y_c = 1 \iff \text{at least one instance of component } c \text{ is present}.$$9 when the prototype count $$\mathbf{L} \in [0,1]^{|\mathcal{C}| \times |\mathcal{C}|},$$0.

5. Architecture, training, and empirical performance in forecasting

Falcon-X formulates forecasting as a masked reconstruction problem over

$$\mathbf{L} \in [0,1]^{|\mathcal{C}| \times |\mathcal{C}|},$$1

After instance-wise normalization, it applies

$$\mathbf{L} \in [0,1]^{|\mathcal{C}| \times |\mathcal{C}|},$$2

where $$\mathbf{L} \in [0,1]^{|\mathcal{C}| \times |\mathcal{C}|},$$3 and $$\mathbf{L} \in [0,1]^{|\mathcal{C}| \times |\mathcal{C}|},$$4 are computed only from observed values. The model augments the input with a relative timestamp sequence

$$\mathbf{L} \in [0,1]^{|\mathcal{C}| \times |\mathcal{C}|},$$5

and a binary observation mask $$\mathbf{L} \in [0,1]^{|\mathcal{C}| \times |\mathcal{C}|},$$6, then tokenizes the triplet with residual patch embedding:

$$\mathbf{L} \in [0,1]^{|\mathcal{C}| \times |\mathcal{C}|},$$7

Temporal encoding is performed independently along time for each variate:

$$\mathbf{L} \in [0,1]^{|\mathcal{C}| \times |\mathcal{C}|},$$8

yielding $$\mathbf{L} \in [0,1]^{|\mathcal{C}| \times |\mathcal{C}|},$$9. Falcon-X then applies Unified Prototype Diff-Attention using two learnable prototype banks $L_{uv}$0. For an entity $L_{uv}$1, the positive and negative affinity maps are

$L_{uv}$2

and the prototype representation is

$L_{uv}$3

After Latent Entity Attention, the model reconstructs variate-specific structure through the Variate Reassembly Router:

$L_{uv}$4

A gated residual fusion then combines temporal and routed cross-variate information:

$L_{uv}$5

The forecasting objective is probabilistic and quantile-based:

$L_{uv}$6

where $L_{uv}$7 is the quantile loss and $L_{uv}$8 penalizes overlap between positive and negative prototype banks. Predictions are mapped back to physical scale via

$L_{uv}$9

Pretraining mixes real and synthetic data from GIFT-Eval, the Chronos training corpus, QuitoBench, TSMixup, KernelSynth, and synthetic multivariate datasets built through grouping and dependency injection. The largest configuration uses 591M parameters, hidden size $u$0, patch length $u$1, 16 Time Attention layers, 16 Entity Attention layers, 16 heads per layer, maximum context length $u$2, and maximum prediction length $u$3. Training uses Megatron-LM, 1,000,000 iterations, global batch size 384, bf16, and AdamW with $u$4, $u$5, weight decay $u$6, warmup to $u$7, and cosine decay to $u$8.

On GIFT-Eval, Falcon-X reports 0.666 MASE and 0.453 CRPS, improving on STRIDE by 1.2% in MASE, on Toto-2.0-FT by 1.9% in MASE and 2.2% in CRPS, and on Timer-S1 by 3.9% in MASE and 6.6% in CRPS. Grouped by horizon, it reports 0.65 MASE in the short term, 0.68 in the medium term, and 0.70 in the long term. On fev-bench, Falcon-X reports 0.652 MASE and 0.490 CRPS, slightly behind Chronos-2 at 0.645 MASE and 0.485 CRPS. Ablations identify the removal of $u$9 as the largest drop, support the gated residual, and place the best prototype count around $v$0 or $v$1. The paper also notes several limitations: sensitivity to the prototype dimension, the need to control cross-variate fusion when dependencies are weak, partial reliance on synthetic multivariate construction, and the fact that the “signed dependence” story remains more architectural than directly interpretable.

6. Misidentification within the broader Falcon/FALCON literature

A persistent source of confusion is that many arXiv papers titled Falcon or FALCON do not define Falcon-X. “Falcon: Fair Active Learning using Multi-armed Bandits” presents a fair active learning framework and explicitly states that the paper never mentions a variant called “Falcon-X” (Tae et al., 2024). “Falcon: A Cross-Modal Evaluation Dataset for Comprehensive Safety Perception” introduces a multimodal safety benchmark named Falcon and an evaluator named FalconEye, again without a Falcon-X variant (Xue et al., 28 Sep 2025). “Falcon: A Remote Sensing Vision-Language Foundation Model” consistently uses Falcon for the model and Falcon_SFT for the instruction-tuning dataset, and does not introduce Falcon-X (Yao et al., 14 Mar 2025).

This suggests that Falcon-X is best treated as a context-dependent label rather than a canonical member of a single named family. In the X-ray literature, Falcon-X is a benchmark paired with a separate model named Falcon. In time-series forecasting, Falcon-X is the model itself. In several other literatures—fair active learning, multimodal safety evaluation, and remote sensing vision-language modeling—the suffix does not appear at all. For rigorous citation and system identification, the surrounding domain and arXiv id are therefore indispensable.