Automating design-to-code translation remains a longstanding goal in software engineering. While Multimodal Large Language Models (MLLMs) have shown promise in design-to-code tasks, existing benchmarks rely solely on screenshots, discarding the structured metadata indispensable in real-world design handoff workflows. Furthermore, current evaluation paradigms are deficient, as code-level metrics fail to account for implementation variability and visual metrics overlook fine-grained defects, while standardized assessments for responsive behaviors remain absent. To address these limitations, we introduce FigmaBench, an industrial-grade benchmark for multimodal design-to-code generation. We collect raw design files from the Figma Community and apply a multi-stage pipeline to curate 1,234 high-quality samples, each comprising high-resolution screenshots and complete JSON metadata. We further propose a comprehensive evaluation framework with four complementary metrics: visual consistency (VCS), structural layout alignment (SLA), textual and stylistic fidelity (TSF), and responsive quality score (RQS). Through extensive evaluation of state-of-the-art MLLMs, we uncover a critical fidelity-responsiveness paradox: models achieving high visual fidelity tend to generate rigid, non-responsive code. We trace this phenomenon to a metadata trap, where models shortcut layout reasoning by directly transcribing absolute coordinates rather than generating fluid structures. Our data and code are available at https://anonymous.4open.science/r/FigmaBench-6C84/.

Exploratory GUI testing is essential for software quality but suffers from high manual costs. While Multi-modal Large Language Model (MLLM) agents excel in navigation, they fail to autonomously discover defects due to two core challenges: Goal-Oriented Masking, where agents prioritize task completion over reporting anomalies, and Execution-Bias Attribution, where system defects are misidentified as agent errors. To address these, we first introduce GUITestBench, the first interactive benchmark for this task, featuring 143 tasks across 26 defects. We then propose GUITester, a multi-agent framework that decouples navigation from verification via two modules: (i) a Planning-Execution Module (PEM) that proactively probes for defects via embedded testing intents, and (ii) a Hierarchical Reflection Module (HRM) that resolves attribution ambiguity through interaction history analysis. GUITester achieves an F1-score of 48.90% (Pass@3) on GUITestBench, outperforming state-of-the-art baselines (33.35%). Our work demonstrates the feasibility of autonomous exploratory testing and provides a robust foundation for future GUI quality assurance.

Large Language Models (LLMs) have achieved strong performance on many code-related tasks, yet they still struggle with repository-level scenarios where reasoning depends on long, noisy, and structurally complex contexts. While existing retrieval methods, including both similarity-based and graph-based approaches, can identify relevant code snippets, they often retrieve excessive contexts that intensify the “lost-in-the-middle” phenomenon and dilute model attention with redundant contexts. To address this, we present RepoDistill, a novel framework that integrates retrieval with learned budget allocation for fine-grained context compression. RepoDistill first employs a plug-and-play lightweight GraphRAG to retrieve context that follows logical flows. It then applies Compression-Aware Budget Allocation guided by Compression-Aware Policy Optimization, which formulates context management as a multi-step decision problem and learns allocation policies for contexts. Experiments show that RepoDistill outperforms baselines, achieving gains of up to +7.00 on SWE-QA, +24.4% on CoderEval, and +0.25 on LongCodeU. Furthermore, a compact 4B-parameter model trained with RepoDistill can serve as an effective context compressor for closed-source LLMs, reducing input tokens by up to 66% while maintaining comparable performance. We release our code at https://anonymous.4open.science/r/RepoDistill-12B0.