arXiv:2606.18218v3 Announce Type: replace-cross Abstract: We study finite-horizon queue peaks in generalized switches, a standard stochastic-network model in which many queues share constrained service resources. Arrivals may be dependent, nonstationary, and responsive to the system history; the only load condition is uniform interior slack, meaning the conditional mean arrival vector stays in a fixed contraction of the capacity region. We show that this slack reshapes the finite-time peak law for drift-minimizing scheduling policies such as MaxWeight. The square-root envelope that is sharp without slack persists only up to a geometry-dependent threshold; beyond that threshold, the running maximum grows only logarithmically with the horizon, both with high probability and in expectation. The mechanism is self-normalization: in the current queue direction, the projected fluctuation scale is normalized by the stabilizing drift scale. This removes capacity geometry from the logarithmic coefficient, while geometry remains in the threshold. Matching lower bounds show that both the logarithmic term and a geometric threshold are unavoidable. When finite-time state-space collapse is available, the threshold can be sharpened using local bottleneck geometry. For generalized input-queued switches, we obtain finite-time peak bounds with tight logarithmic coefficients. Simulations illustrate the two-phase envelope, local geometric refinements, and variance-sensitive improvements predicted by the theory.
Science Journals
arXiv:2607.00836v3 Announce Type: replace Abstract: World models are increasingly used in embodied intelligence and generative simulation, yet their scope remains ambiguous across communities. This tutorial presents a design-space view of world models as action-conditioned predictive models that estimate the future evolution of task-relevant observations or states. We categorize existing methods into observation-space and state-space world models, comparing their trade-offs in visual fidelity, spatial structure, physical interpretability, and control usability. We further introduce world action models, which connect predicted futures with executable robot actions, and summarize four representative paradigms: imagine-then-execute, video-feature-conditioned action prediction, joint video-action modeling, and auxiliary video prediction for policy learning. The goal of this tutorial is to clarify the conceptual scope of world (action) models and provide a structured taxonomy for embodied prediction and control.
arXiv:2607.04151v1 Announce Type: new Abstract: Modern smartphones capture Live Photos, short video bursts surrounding a still image, offering a dynamic and engaging photographic experience. However, the cover photo and video components are generated by two distinct imaging pipelines: the photo stream undergoes full computational photography processing, while the video stream is constrained by real-time efficiency and heavy compression. This intrinsic separation produces a substantial quality gap in resolution, color fidelity, and dynamic range between the cover photo and video frames. When users reselect an alternative frame from the video to replace an imperfect cover, the chosen frame often suffers from severe degradation, making direct replacement visually unsatisfactory. Restoring such frames requires simultaneous enhancement of spatial detail and color appearance, a task considerably more challenging than ordinary super-resolution or color enhancement. To address this, we define the Live Photo Cover Frame Reselection and Enhancement (LPRE) task, which leverages the intrinsic cues available within each Live Photo: the high-quality cover image as a structural and color reference, the user-reselected low-quality frame as the reconstruction target and several adjacent video frames providing temporal cues. Building upon this formulation, we construct Live2K, a real-world dataset of 2,042 Live Photos, and develop a unified one-stage baseline that integrates multi-frame fusion, guided color enhancement and super-resolution, establishing the first benchmark for Live Photo enhancement research.
arXiv:2607.04181v1 Announce Type: new Abstract: Online large language model (LLM) serving has become the backbone of modern AI applications, powering diverse downstream services through shared hardware clusters. However, modern serving systems frequently encounter highly dynamic workloads characterized by severe workload skewness, where a small fraction of model instances receives the vast majority of traffic. Existing instance-level scaling mechanisms are limited by coarse-grained resource adjustment: scaling up requires the cold-start of full-model replicas, incurring substantial latency, while scaling down leaves the system vulnerable to performance degradation during sudden traffic surges. The key insight of this work is that LLM serving offers a unique opportunity for fine-grained scaling. In this paper, we propose CoCoScale, a layer-wise dynamic scaling mechanism that selectively expands the parallelism of hot layers onto idle resources reclaimed from underutilized devices, enabling elastic data parallelism without altering model architectures or adding hardware overhead. Evaluations demonstrate that CoCoScale significantly reduces cold start latency by 97.9%-99.3% compared to traditional scale up. Under production traces, CoCoScale reduces average latency by 20.7\%--28.1\% and achieves full Service Level Objective (SLO) attainment, demonstrating superior dynamic adaptability and resource efficiency.
arXiv:2607.04287v1 Announce Type: new Abstract: The Weighted Center} problem takes as its input a graph $G=(V,E)$ together with a profile $\pi$ such that every vertex $v$ is mapped to some nonnegative multiplicative weight $\pi(v)$. Its output must be some vertex $c$ minimizing $\max\{\pi(v)d_G(c,v) : v \in V\}$. The classic Center problem corresponds to the case where $\pi(v) =1$ for every vertex $v$. In the literature, various almost linear-time algorithms have been proposed for the Center problem on some well-structured classes of graphs. By contrast, similarly efficient algorithms for the Weighted Center problem have been scarce. We investigate how the Gromov hyperbolicity, alone or in combination with other metric and geometric properties on graphs, can be used in the design of exact and approximate almost linear-time algorithms for the Weighted Center problem. In particular, we derive almost optimal algorithms for the following well-studied classes of graphs: chordal graphs, distance-hereditary graphs (both in $\mathcal{O}(m)$ time), dually chordal graphs and chordal bipartite graphs (both in $\mathcal{O}(m\log{n})$ time).
arXiv:2607.04325v1 Announce Type: new Abstract: REST APIs are widely used in industry. Therefore, a lot of research has been focused on how to automatically generate test cases for REST APIs, with few different open-source fuzzers existing in the literature. For a thorough testing, especially in black-box scenarios, just relying on the information provided in the OpenAPI schemas is not enough. Testers typically need to provide extra input data to help steer the fuzzers in the right direction. Dedicated formats specific to each different fuzzer would work, but they would create a vendor lock-in, as well as increasing cognitive load. The OpenAPI Initiative (OAI) standard Overlay might be a solution to this problem. Such standard enables to define transformations on the OpenAPI schemas, where testers can provide input data in Overlay files where such data is provided as ``examples'' entries. In this paper, we have extended the state-of-the-art fuzzer EvoMaster to support Overlay files natively. Experiments are carried out in industry on five APIs from five enterprises from around the world (e.g., Belgium, China, Germany and T\"urkiye), including two Fortune500 enterprises as well as a 3-man startup. Our industrial results show that Overlay is a viable solution to better enable black-box fuzzing of REST APIs in industry.
arXiv:2607.04329v1 Announce Type: new Abstract: Large language models increasingly operate in settings where humans are active collaborators rather than passive task providers. We introduce HAS-Framework, a graph-based framework that represents humans and LLM-powered agents as first-class participants with explicit roles, permissions, communication paths, and action authority. Building on this framework, HAS-Bench evaluates Human-Agent Systems under configurable human participation across agency levels, interaction channels, and persona policies. The benchmark measures both task outcomes and process-level collaboration behavior, including clarification quality, feedback utilization, control calibration, safety, initiative, and interaction cost. Experiments across six domains show that human participation can substantially improve task completion and failure recovery, but the gains depend on when, how, and by whom human input is exercised.
arXiv:2607.04636v1 Announce Type: new Abstract: Key Information Extraction (KIE) converts visually rich documents into structured data, but practical deployment remains challenging: strong performance often relies on costly on-server Large Multimodal Models (LMMs), while compact locally deployable models lack sufficient KIE supervision. We present SAYRE, a scene-aware document synthesis framework for generating scalable KIE training data without hand-crafted template design. Given a few exemplar documents, SAYRE captures category-specific content patterns and layout conventions to synthesize document-schema-annotation triples. It further introduces error-driven generation, which expands real-world failure cases into hard training examples while preserving their structural patterns. Experiments on constrained- and open-category KIE show that SAYRE consistently improves Qwen3-VL backbones and achieves the strongest overall performance among on-device LMMs. Data scaling experiments show an overall upward trend as more synthesized data is introduced, especially for smaller models and open-category extraction. Error analysis further shows that synthesized training reduces field-level errors by improving schema-aware extraction over dense tables, business identifiers, and contract clauses. These results establish scene-aware synthesis as an effective data-centric approach for improving practical multimodal KIE.
arXiv:2607.04637v1 Announce Type: new Abstract: Vision-Language-Action Models (VLAs), which leverage the advanced reasoning capabilities of Vision-Language Models (VLMs), show promising generalization in complex autonomous driving scenarios. Existing VLAs typically predict and optimize 3D trajectories from 2D images. While intuitive, this 2D-to-3D prediction is inherently entangled with camera parameters, leading to limited data scalability across heterogeneous driving datasets. Moreover, directly optimizing in 3D space induces severe convergence to trivial solutions, where VLAs rely on ego-status rather than visual scene understanding. To address these issues, we propose PixelPilot, a novel VLA featuring a decoupled planning and lifting paradigm. In the planning phase, PixelPilot reformulates scene understanding and trajectory prediction as sensor-agnostic 2D-to-2D tasks in the image plane, thereby facilitating scalable training across diverse datasets. The planned 2D trajectories are then deterministically lifted to 3D only during inference, ensuring the full exploitation of visual cues and generalization across different vehicles. To realize this paradigm, we propose a knowledge-instilled policy learning strategy that applies dense, intermediate rewards via Group Relative Policy Optimization (GRPO) to enforce a rigorous causal chain from visual perception to spatial planning. Extensive experiments demonstrate that PixelPilot achieves state-of-the-art performance in both open-loop and closed-loop settings, validating its superior scalability and visual reasoning capabilities.
arXiv:2607.05105v1 Announce Type: new Abstract: We study a Marker-and-Cell (MAC) scheme for the barotropic Euler system. First, we apply the recently developed Lax-type convergence theorem to show the convergence of the MAC scheme to i) a dissipative weak solution unconditionally and ii) a strong solution as long as it exists. Second, We derive relative energy error estimates up to the lifespan of a strong solution, without assuming uniform boundedness of the numerical sequence. Additionally, assuming the boundedness of the numerical solutions, we obtain the optimal relative energy rate of 1, corresponding to a convergence rate of 1/2 for the numerical solutions. Finally, we corroborate our theoretical results by numerical experiments.
arXiv:2607.04304v1 Announce Type: new Abstract: Traditional semantic segmentation models operate under a closed-set assumption and struggle to recognize unknown or unexpected objects-an essential capability for autonomous driving. As a result, such models often misclassify or overlook out-of-distribution (OOD) road anomalies, posing safety risks in open-world environments. We present a lightweight, postprocessing, road-aware anomaly segmentation framework that requires no retraining, no OOD data, and no auxiliary supervision. Our approach builds on a mask transformer-based segmentation network by exploiting query-level mask confidence and deriving a polygonal road prior to detect gap regions that may correspond to anomalies. To further suppress false positives, we introduce a CLIP-based zero-shot semantic filtering module using in-distribution prompts, with optional generalized OOD prompts. By jointly leveraging spatial priors and semantic verification, our framework produces robust and interpretable anomaly predictions. Evaluation on three public benchmarks-Fishyscapes, SMIYC, and RoadAnomaly-shows consistently strong performance. In particular, our method outperforms the training-free baseline Maskomaly on most metrics and achieves the highest AP on Fishyscapes LostAndFound. These results demonstrate the practicality and deployability of our approach for real-world autonomous driving systems.
arXiv:2607.04872v1 Announce Type: new Abstract: Reasoning temporal localization (RTL) requires a model to generate an answer that itself contains the time interval supporting it, so high-level reasoning and precise temporal grounding must be produced jointly in a single response. To tackle this challenging task, we propose the first event-centric video chain-of-thought framework, dubbed EventCoT. EventCoT first performs event-centric tokenization of the input video to convert it into compact event tokens, enabling efficient identification of question-relevant events. It then reasons within the identified events to generate the answer, grounding the time interval via embedding matching that aligns placeholder tokens with visual embeddings. EventCoT achieves state-of-the-art results on ActivityNet-RTL for reasoning temporal localization while using substantially fewer visual tokens than previous work. To verify its general performance, we further evaluate EventCoT on the grounded video question answering benchmark ReXTime, where it attains strong zero-shot results.
arXiv:2607.04880v1 Announce Type: new Abstract: Recent advances in large-scale pretrained vision-language-action models have improved robot policy learning, but directly deploying such policies in user-specific environments remains challenging due to limited generalization, which inevitably requires collecting a dataset tailored to the target environment. Teleoperation yields well-aligned data but is costly and difficult to scale, whereas simulation scales easily but struggles to resemble the target environment and generate task-specific trajectories. To meet both simultaneously, we propose PRISM, an end-to-end pipeline that generates personalized robotic datasets from a single image and a natural-language instruction. PRISM constructs digital cousin scenes that are semantically and geometrically aligned with the user environment yet diverse at the instance level, and synthesizes executable demonstrations without human teleoperation. Extensive experiments show that policies trained on PRISM-generated datasets outperform those trained on baseline-generated datasets on LIBERO and LIBERO-Plus, achieve up to 100\% success rate on three real-world manipulation tasks, and maintain stronger performance when evaluated in environments that differ from those seen during training.
arXiv:2607.04500v1 Announce Type: new Abstract: Self-supervised latent world models can assign a surprise score to driving scenarios without any human labels. A natural follow-up question is whether such a model, trained on driving data from one geographic region, can generalize its notion of complexity to unseen cities and sensor configurations. We study this question through a controlled transfer experiment: we train JEPA-based world models on nuPlan data (Pittsburgh, Boston, Singapore) and evaluate zero-shot on held-out Argoverse 2 validation scenarios from Miami and Austin. We find that models trained on geographically diverse data generalize significantly better than models trained on equal amounts of single-geography data. In a matched-scale ablation at 63,000 scenarios per condition (n=3 seeds each), combined training reduces mean surprise score by 16.5% relative to nuPlan-only training (0.228 +/- 0.015 vs 0.273 +/- 0.008). Notably, training on 200,000 AV2-only scenarios (3x more data from one geography) still produces higher surprise (0.264) than the combined 63K model, suggesting that geographic diversity is a stronger predictor of cross-domain generalization than raw data volume.
arXiv:2607.04890v1 Announce Type: new Abstract: Addressing hate and violence in society requires timely detection of hateful events from public reporting, but automated identification of hateful events remains underexplored. We introduce the task of hateful event detection and investigate the ability of AI systems, specifically large language models (LLMs), to discover and classify reports of antisemitic events with fine-grained labels. We evaluate OpenAI's GPT-4o and Meta's Llama-3.2-3B-Instruct on multiple expert-annotated datasets containing antisemitic event descriptions from news articles, civil society reports, and official records. We show that LLMs, particularly GPT-4o, have potential for this task, but substantial improvement is needed. Providing clear term definitions and in-context examples in prompts can improve performance: definitions are most helpful for rhetoric-oriented events (e.g. classical antisemitic tropes), while examples help label action-oriented events (e.g. physical assault). A case study of college newspapers demonstrates that LLMs can help surface relevant real-world events, supporting early monitoring and intervention. Overall, our findings highlight both opportunities and critical gaps in AI's ability to recognize complex harms and underscore the need for collaborative efforts among AI developers, policymakers, and civil society to design models, implement robust evaluation, and develop policy frameworks for defining and combating hate efficiently and effectively.
arXiv:2607.04948v1 Announce Type: new Abstract: Integrating AI agents into Software Engineering (SE) raises an important challenge: how can we specify and realize AI agents that work effectively alongside humans in hybrid SE teams? Determining the right granularity and separation of concerns for such agents is non-trivial. Coarse-grained agents may introduce unmanageable complexity, whereas micro-agents may create severe coordination overhead. Moreover, existing multi-agent SE frameworks typically rely on predefined role structures and do not account for project-specific characteristics or process adaptations. We address this by combining object-centric, imperative, and declarative process mining. Using event logs extracted from software repositories, our approach discovers project-specific agent roles using a predefined SE role vocabulary grounded in repository behavior and generates matching agent specifications and implementations. As proof-of-concept, we applied our approach to a well-established open-source project. We performed functional tests and an exploratory user study to determine how well the generated AI agent specifications are aligned with human expectations.
arXiv:2607.04548v1 Announce Type: new Abstract: Novel category discovery aims to identify unseen classes from unlabeled data by transferring knowledge from labeled categories, but most existing methods perform discovery in opaque latent feature spaces. As a result, they may separate novel categories accurately while providing little insight into what semantic evidence defines each discovered group. We propose xNCD, an explainable novel category discovery framework that performs both representation-based discovery and pseudo-label assignment directly in a structured semantic concept space. Instead of clustering arbitrary deep features, xNCD learns a label-free concept representation by aligning visual features with vision-language similarity priors from pretrained multimodal models, and then applies a unified labeled-and-unlabeled self-labeling objective over concept-space logits. This design makes each discovered category explainable by construction through stable concept signatures and instance-level concept evidence. Theoretically, we show that routing discovery through a semantic concept bottleneck induces a strict restriction of the feature-space hypothesis class, excluding a large family of unconstrained decision rules and biasing induced partitions toward semantically interpretable concept coordinates. Experiments on CIFAR-10, CIFAR-100, and CUB-200 demonstrate that xNCD preserves strong discovery performance while providing intrinsic explanations. Under task-agnostic evaluation, xNCD achieves 92.63% overall accuracy on CIFAR-10, close to UNO's 93.4%, and improves CIFAR-100 overall accuracy from 73.2% to 76.45%, while being the only compared method that provides human-readable cluster- and instance-level explanations.
arXiv:2607.05142v1 Announce Type: new Abstract: Standard Pancharatnam-Berry and linear-birefringent media convert optical spin into orbital angular momentum (OAM) through an anisotropy \emph{director}, a rank-two, headless field, and therefore obey the selection rule $\Delta\ell=2q$ per unit texture charge $q$. We show that a medium with geometric \emph{torsion}, the continuum limit of a screw-dislocation array, can convert spin to OAM through the \emph{contortion} of its material connection, which enters the effective paraxial dynamics as a rank-one vector field. The resulting selection rule is $\Delta\ell=q$. Its winding is fixed by geometry and symmetry, not by a Pancharatnam--Berry director, and the process conserves the screw charge $\tilde J_z=L_z+(q/2)\sigma_z$ while exchanging $(2-q)\hbar$ of angular momentum per converted photon with the defect lattice. Paraxial simulations confirm the rule: a circular Gaussian input develops a stable, topologically quantized $\ell=+q$ vortex in the reversed helicity, with $83\%$ conversion over three Rayleigh ranges and no fine-tuning. We propose a polarization-resolved photonic-lattice discriminator in which the slope of the measured OAM versus the independently written texture charge, one for torsion, two for birefringence, separates the two mechanisms.
arXiv:2607.05227v1 Announce Type: new Abstract: The reliance on software as a distinguishing factor in the automotive industry is increasing. With a combined reliance on vendor-supplied software and cost-effective implementation, the AUTOSAR consortium was initialized to provide standardized platform specifications that enable re-use. Specifically, the AUTOSAR Adaptive Platform (AP) specification aims to provide a high-performance service-oriented architecture. Objective: The goal of this study is to investigate what pain-points emerge when developing AUTOSAR Adaptive applications and whether they originate from the platform specification, its vendor-implementation, or its local usage. Methods: We conduct a Design Science Research study, developing a minimal AP that serves as an experimental prototype for our investigation. Results: We find that a combination of specification-inherent, implementation-based, and local practices contributes to the emergence of pain-points. Conclusions: We conclude that there are AUTOSAR specification-inherent reasons for pain-points, resulting from architectural choices and re-use goals. The implication for development organizations is the need to mitigate these effects through tooling that better supports configuration file management and reduces developer training time to properly understand the adaptive application runtime life-cycle.
arXiv:2607.04105v1 Announce Type: new Abstract: The curve shortening--DeTurck flow introduced by Elliott and Fritz (IMA J. Numer. Anal. 37(2): 543--603, 2017) is a reparameterization of the curve shortening flow via the DeTurck trick. For corresponding discrete schemes, although the $H^1$-error estimate has been established, the optimal $L^2$-error estimate remains open due to technical difficulties. In this paper, we prove optimal $L^2$-error estimates for linearized Euler and Crank--Nicolson discretizations combined with finite elements of degree $k\ge 1$ in space. Moreover, we provide an extrinsic approach to derive the curve shortening--DeTurck flow. Numerical experiments are presented to illustrate the mesh distribution properties and to verify the convergence rates in both space and time.
arXiv:2607.05263v1 Announce Type: new Abstract: This paper concerns the problem of unsupervised temporal action segmentation for long, untrimmed videos. Recent successful approaches follow a joint representation learning and clustering paradigm, where optimal transport (OT) is adopted to produce pseudo labels for learning frame representations. These approaches alternate between estimating pseudo labels using OT and optimizing the parameters with gradient descent during training, where OT is used for obtaining the final temporal action segmentation. A major limitation of these works is that they learn a deterministic embedding for frame representations. The iterative procedure between learning deterministic embeddings based on pseudo labels and estimating pseudo labels from the learned embedding can thus get quickly stuck in a local optimum. As an alternative, we thus propose to learn a probabilistic embedding for frame representations. The embeddings are modeled by Gaussian distributions and we sample from the distributions before estimating the pseudo labels. We evaluate our approach on several challenging temporal action segmentation datasets and achieve results comparable to, and in some cases, better than the state of the art. Compared to baselines with deterministic embeddings, our approach improves MoF up to 20.7\% and F1-score up to 19.0\%. Our code is available at https://github.com/derkbreeze/PEOT.
arXiv:2607.04609v1 Announce Type: new Abstract: Vision-Language-Action (VLA) policies that execute fixed-length action chunks can exhibit multimodal bifurcation: a cross-chunk inconsistency in which adjacent chunks generated from independent Gaussian latents can converge to incompatible trajectory modes, producing abrupt discontinuities at chunk boundaries. Existing remedies either require backpropagation through the policy at each denoising step, rely on rejection sampling, or require retraining, each trading computational cost or task reliability for smoother transitions. We propose SEAM (Smooth Execution of Action-Chunked Motion), a training-free inference-time method for flow matching VLAs. SEAM exploits a simple synchronous-execution insight: after the robot consumes the executed prefix, the previous chunk's unexecuted tail is already available as an analytic consistency reference. Its core mechanism, Velocity-guided Loss Steering (VLS), derives a time-dependent target from this tail and applies a closed-form correction after each Euler step without backpropagating through the policy network. On LIBERO-10 with pi_0.5, SEAM reduces boundary jerk by 28%, reduces chunk transition discontinuity by 27%, preserves baseline-level task success, and keeps denoising-loop cost near the unguided baseline.
arXiv:2607.05347v1 Announce Type: new Abstract: While feedforward 3D reconstruction excels at efficient novel view synthesis, it typically falters when faced with scenes under varying illumination. To this end, we introduce WildSplat, the first feedforward 3D Gaussian Splatting framework capable of appearance-conditioned novel-view synthesis for unposed in-the-wild images. To handle inconsistent photometric conditions, we propose a dual-branch architecture that explicitly decouples geometry from appearance. The geometry branch extracts an appearance-invariant 3D structure and jointly predicts camera poses. To govern the rendering appearance, the appearance branch injects target appearance cues into the content features via a globally pre-modulated cross-attention mechanism. To further prevent feature entanglement, we introduce a joint multi-reference training strategy that stabilizes the training process. Extensive experiments show that WildSplat surpasses existing optimization-based and feedforward methods, achieving state-of-the-art performance in in-the-wild novel view synthesis and appearance editing from sparse inputs in a single forward pass.
arXiv:2607.04379v1 Announce Type: new Abstract: This paper presents an adversarial security study of the Policy-Aware LLM Retrieval-Augmented Generation (PA-LLM-RAG) framework for Internet of Battlefield Things (IoBT) mission control. We propose Query-Agnostic Semantic Retrieval Poisoning, a novel attack that injects semantically crafted rules into the IoBT knowledge base achieving high retrieval ranking across all operator query types without requiring knowledge of runtime prompts. The attack achieves 85% LLM context corruption from a single injected rule (1.6% poisoning rate) and saturates at 7.7% poisoning, demonstrating that even minimal knowledge base compromise is sufficient to corrupt mission decisions. To counter this threat, we propose CLD-KB (Cyber-Layered Defense for Knowledge Base), a dual-detector anomaly detection framework combining One-Class SVM boundary detection with a novel Member-Based Category Spread analysis that exploits the three-category IoBT policy taxonomy to identify poisoned rules before they reach the decision LLM. CLD-KB significantly outperforms five baseline methods including DBSCAN, LOF, K-Means, Isolation Forest, and One-Class SVM in both poisoning detection and knowledge preservation, with only 7ms computational overhead per mission, establishing it as an effective and edge-deployable defense for LLM-driven IoBT mission systems.
arXiv:2607.04547v1 Announce Type: new Abstract: People respond to artificial intelligence chatbots (AICs) in highly variable ways. In this paper, we adapt Bronfenbrenner's theory into a heuristic framework for understanding this variation. The framework places the human user at the center while also placing the AI there and reconceptualizing the proximal processes as the repeated, reciprocal, and coadaptive interactions between the user and a personalized AIC. The surrounding systems identify the contextual factors that shape how the user experiences, interprets, responds to, and is changed by these interactions. Because stateful AICs learn from accumulated exchanges with their users and have memory, users are responding not only to an AIC but also to a version of the AIC that their own prior interactions have helped create. This extension preserves Bronfenbrenner's emphasis on proximal processes while accounting for the unique dynamics of personalized AICs. The resulting framework provides a structured map of where and how variation in human and AIC relationships arises, as well as having implications for researchers, practitioners, and AIC designers.