design intermediate representations

Downstream probing only assesses task-relevant information in representations, failing to characterize critical properties—such as equivariance, invariance, and disentanglement—that govern interpretability and generalization; moreover, existing evaluation frameworks lack standardization, modularity, and cross-modal applicability. Method: We propose the first representation quality assessment framework that transcends downstream tasks, employing controlled factorial probe design to systematically quantify informativeness, equivariance, invariance, and disentanglement. The framework is modular, interpretable, and supports cross-modal analysis (e.g., image and speech). Contribution/Results: It establishes the first standardized, multi-dimensional semantic attribute disentanglement protocol. Experiments reveal substantial divergence in intrinsic representation properties—even among models with comparable downstream performance—enabling fine-grained representation understanding, diagnosis, and optimization. This work introduces a novel paradigm and practical toolkit for representation evaluation beyond task-specific metrics.

The absence of a mathematically formalized representation of design spaces renders design decisions heavily experience-dependent, hindering the development of automated support. Method: This paper introduces an orthogonal discretization model for design spaces—establishing the first structured spatial representation—and integrates, for the first time, large language model (LLM)-driven constraint generation with Monte Carlo tree search (MCTS) to enable autonomous, efficient exploration. It further develops a domain-adaptive instantiation engine that maps abstract design decisions to concrete implementations. Contribution/Results: The framework exhibits cross-domain transferability. Empirical evaluation on data article generation and chart visualization tasks demonstrates significant performance gains over baselines. User studies and expert interviews confirm its effectiveness, usability, and measurable improvement in design quality.

Neural network internal representations often lack stability and cross-architectural consistency due to architectural disparities, hindering knowledge transfer and modular deployment. To address this, we propose a structured regularization framework comprising linear shaping operators and rectified path constraints, which explicitly encode inductive biases to improve geometric alignment of representations across architectures. Through theoretical analysis, controlled transfer experiments, and a novel representation alignment metric, we systematically demonstrate that structural priors significantly enhance semantic consistency among heterogeneous models. Our method improves downstream task performance in model distillation and modular learning by up to 12.3%, offering an interpretable and scalable paradigm for building robust, composable deep learning systems.

Existing CAD generation methods struggle to simultaneously preserve modeling history, topological reference stability, and feature-level editability in cross-platform scenarios. This work proposes CADIR—an agent-oriented, executable intermediate representation that explicitly constructs a procedural graph encompassing operation sequences, parameter dependencies, constraints, and topological selections based on the OpenCASCADE (OCCT) geometric kernel. To enable faithful cross-platform model reconstruction, CADIR introduces a geometric signature matching mechanism. It is the first approach to support explicit procedural graph representations that allow editing across heterogeneous CAD backends. By integrating text- or image-driven procedural graph retrieval, CADIR demonstrates high-fidelity, editable reuse of complete models and substructures across FreeCAD, SolidWorks, and Fusion 360, enabling seamless subsequent modifications.

Existing generative 3D modeling approaches rely on iterative regeneration, which suffers from limited controllability and unpredictable outcomes. This work proposes a compositional modeling workflow that introduces, for the first time, a part-level mixing mechanism supporting hybrid 2D/3D interaction: users assemble 3D models by selecting 2D regions or 3D mesh fragments from multiple candidate generations. The system then automatically synthesizes geometrically consistent models aligned with high-level user intent through a pipeline integrating text- or image-based guidance, geometric segmentation, seamless fusion, and detail refinement. User studies demonstrate that this approach significantly outperforms conventional regeneration paradigms in terms of creative control, alignment with user intent, and overall user satisfaction.

This study addresses the prevailing gap in AI education, which emphasizes model development while neglecting system engineering practices, leaving students ill-equipped to handle real-world challenges such as architectural design, deployment, and monitoring. To bridge this gap, the authors implemented a master’s-level course in which students built a movie recommendation system under realistic constraints, with a focus on integrating AI components into robust software systems, adopting data-driven machine learning practices, and cultivating systems-level thinking. Using a mixed-methods approach—combining analysis of student project artifacts with survey data—the research evaluates learners’ performance in architectural decision-making, integration of heterogeneous models, and adaptation to evolving requirements. Findings reveal common difficulties students encounter in AI system engineering and demonstrate the course’s effectiveness in addressing critical deficiencies in AI engineering education and enhancing systems-aware competencies.

This work addresses the prevailing lack of systematic understanding of foundational formal theories in current AI compiler design, which hinders rigorous evaluation of the completeness and desirability of intermediate representations and compilation abstractions. For the first time, it systematically establishes precise correspondences between core mechanisms of MLIR—such as term rewriting systems, refinement calculi, and abstract interpretation—and classical formal theories. By grounding compiler abstractions in formal semantics, the paper clarifies the theoretical underpinnings of these constructs, articulates a precise notion of “design completeness,” and provides assessable criteria and guiding principles to navigate trade-offs between engineering pragmatism and theoretical ideals.

This work addresses the limitation that conceptual knowledge in large language models (LLMs) is implicitly encoded through statistical correlations, lacking explicit, structured, and composable representations—hindering model stability, controllability, and alignment with human cognition. To tackle this, the paper introduces the first design-space taxonomy for conceptual structures in LLMs, proposing a “four-stage–two-source” framework that spans training, architecture, inference, and interpretation, combined with internal derivation and external anchoring. The framework systematically integrates probing, dictionary learning, and external knowledge-guided approaches, revealing critical gaps in current research—particularly insufficient exploration of the inference stage, fragmentation across stages, and inconsistent terminology. By shifting the paradigm from passive discovery to active design of conceptual representations, this study provides a theoretical foundation and strategic direction for developing LLMs that are more controllable, interpretable, and cognitively aligned.