design space exploration

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.

Exploring complex systems faces the “constrained diversity” challenge—arising from vast parameter spaces, highly nonlinear parameter-to-pattern mappings, and users’ prior expectations about specific emergent patterns. Method: We propose a “constrained diversity” framework that integrates user-defined explicit constraints with interactive human-in-the-loop collaboration to maximize pattern diversity efficiently within user-specified regions while preserving global coverage. Our approach innovatively incorporates human guidance into active sampling, jointly modeling nonlinear parameter–pattern mappings and enforcing constraint satisfaction through iterative feedback. Contribution/Results: The framework supports system-agnostic constraint specification and sample-efficient exploration. Experiments across diverse complex systems demonstrate significant improvements in both diversity and efficiency of pattern discovery within target regions, effectively balancing localized focus with global exploration.

Addressing three key challenges in CPU microarchitecture design space exploration (DSE)—degraded surrogate model accuracy and scalability, inefficient acquisition strategies, and poor interpretability—this paper proposes AttentionDSE, the first DSE framework integrating attention mechanisms. It unifies high-accuracy performance prediction with real-time bottleneck identification via interpretable, dynamic mapping from architectural parameters to performance contributions. Methodologically, AttentionDSE synergistically combines attention-based modeling, multi-objective optimization, and Pareto frontier search, enabling adaptive analysis under design modifications. Evaluated on the SPEC CPU 2017 benchmark suite, it reduces exploration time by over 80% and improves Pareto hypervolume by 3.9%, while significantly outperforming state-of-the-art approaches in both prediction accuracy and scalability.

High computational cost, slow convergence, and suboptimal results in high-level synthesis (HLS) design space exploration (DSE) stem from combinatorial explosion of optimization directives. Method: This paper proposes the first large language model (LLM)-driven DSE framework, which leverages LLMs’ semantic understanding of hardware design quality to enable intelligent pruning and calibrated initial sampling. The framework synergistically balances convergent and divergent thinking to jointly optimize solution quality and diversity. Contribution/Results: Compared to conventional heuristic methods, the framework achieves 5.1×–16.6× improvement in approximating the reference Pareto front. It attains equivalent multi-objective optimization performance using only 4.6% of the exploration overhead required by NSGA-II, significantly enhancing both efficiency and accuracy of multi-objective DSE in HLS.

Traditional design space exploration (DSE) for computer architectures relies heavily on black-box simulation and lacks the reasoning capability of human experts who integrate physical constraints with workload structure. This work proposes a closed-loop simulation framework powered by a general-purpose large language model (LLM) agent, which—without any fine-tuning—automatically guides architectural DSE by emulating human architects’ reasoning processes. By integrating an LLM agent, closed-loop interaction with a simulator, automated code generation, and an interpretable architectural reasoning mechanism, the approach achieves comparable or superior design quality in tasks such as DNN accelerator mapping, hardware-software co-design, and CPU cache optimization, using only 1%–2% of the simulation budget required by conventional methods. Moreover, it produces traceable and auditable decision trajectories.

This work addresses the challenges of design space exploration (DSE) for GPU architectures targeting modern AI workloads such as large language model (LLM) inference, which involve an enormous search space, high simulation costs, and complex multi-objective optimization. For the first time, LLMs are integrated into GPU architecture exploration to automatically extract architectural knowledge from simulator code, enabling the construction and dynamic refinement of design rules. These rules, combined with sensitivity analysis and bottleneck identification, efficiently guide the optimization process. The study introduces the first DSE benchmark featuring three core capability evaluations and implements a mechanism for automatic rule generation and correction. Within a design space of 4.7 million configurations, the approach identifies six designs surpassing the NVIDIA A100 in just 20 exploration steps, achieving a 17.5× improvement in exploration efficiency and a 32.9% gain in Pareto hypervolume over baseline methods.

This work addresses the inefficiency of conventional black-box optimization in hierarchical heterogeneous search spaces where a large fraction of configurations are infeasible due to crash risks. To tackle this challenge, the authors propose a feasibility-first exploration strategy combined with a novel Thermal Budget Annealing (TBA) method that decomposes optimization into distinct exploration and exploitation phases. Robustness is further enhanced through trial timeouts, subspace blacklisting, and Tree-structured Parzen Estimators (TPE). The study introduces DeployBench, a new benchmark featuring hidden crash zones and non-uniform evaluation costs. Experimental results demonstrate that the proposed approach significantly improves the success rate of discovering valid model families under constraints and substantially reduces the overhead of invalid evaluations in both synthetic and real-world GPU deployment tasks.

This work addresses the mismatch between existing AI-powered design tools—which rely on static or continuously expanding design spaces—and the dynamic needs of designers, who require broad exploration followed by focused refinement. To bridge this gap, the authors propose an axis-centered workflow that dynamically expands and contracts the design space to support structured exploration and refinement. They implement this approach in a prototype system, Surprise2Refine, introducing a novel axis-centric interaction paradigm that enables the design space to adaptively evolve according to creative intent. This enhances users’ sense of control and traceability throughout the creative scaffolding process. A controlled study with 14 designers demonstrates that the system significantly improves perceived user control, facilitates tracing of scaffolding pathways, and yields design outcomes rated as more creatively compelling.

This work addresses the limitations of existing image exploration tools, which overly rely on similarity-based ranking and thereby constrain designers’ holistic perception of visual space and pattern discovery during early-stage ideation. The authors propose an interactive exploration prototype that supports gradual adjustment of diversity, introducing for the first time a dynamic interface for explicitly negotiating the trade-off between diversity and similarity—departing from conventional static ranking paradigms. Built upon Determinantal Point Processes (DPPs), the system enables controllable diversity-aware sampling and exposes the underlying tuning mechanism through an intuitive interface. User studies demonstrate that this approach significantly reduces backtracking behavior, facilitates visual discovery, and outperforms existing baseline tools during the initial phases of creative exploration.