Large Language Models (LLMs) can translate natural language requirements into code, yet empirical analyses of representative models reveal that semantic errors—programs that compile but behave incorrectly—constitute the majority of observed faults (e.g., >60% on DeepSeek-Coder-6.7B and QwenCoder-7B). Post-hoc repair pipelines detect such faults only after execution, incurring latency, relying on incomplete test suites, and often mis-localizing the defect. Since semantic drift originates in the autoregressive decoding process, intervening while the code is being generated is a direct way to stop error propagation. Constrained-decoding approaches such as ROCODE attempt this, but still wait until the entire program runs to obtain feedback and use entropy heuristics that do not truly capture semantics. A more effective solution must inject semantic signals—early and precisely—into the decoding process. We present SemGuard, a semantic-evaluator-driven framework that performs real-time, line-level semantic supervision. To train the evaluator, we build SemDiff, the first dataset with fine-grained annotations that mark the exact line where a correct and an incorrect implementation diverge. The evaluator, once embedded in the LLM’s decoder, flags deviations on partial code, rolls back to the faulty line, and guides regeneration—without executing the program or requiring test cases. Across four benchmarks, SemGuard consistently outperforms state-of-the-art baselines. It lowers the semantic error rate by 19.86% on SemDiff relative to ROCODE, and lifts Pass@1 by 48.92% on the realworld LiveCodeBench with CodeLlama-7B. Similar gains hold for StarCoder2-7B on MBPP and for DeepSeekCoder-6.7B on the Java benchmark SemDiff-Java, demonstrating model- and language-agnostic effectiveness.

Multi-label image classification poses a formidable challenge due to the presence of multiple objects in each image, rendering it notably complex to decipher the visual content comprehensively. Discriminating between multiple objects necessitates the establishment of robust visual label dependencies. Previous methods attempt to formulate cross-modal interaction or one-shot co-occurrence relationship guidance. However, it not only exhibits limitations when handling occluded or blurry objects but also fails to fully leverage the diverse hierarchical properties for sustainably guiding the learning process of label dependencies. To sustainably establish hierarchical visual label dependencies, this paper introduces a Pyramidal Cross-modal Transformer framework for MLIC tasks. Specifically, the pyramidal visual guidance layer parses the visual features into a multi-resolution pyramid structure, allowing the updated visual-related information to provide sustained guidance for label semantics. This surpasses the conventional pre-processing of co-occurrence relationships. Besides, the hybrid modal interaction layer is proposed to effectively mitigate the semantic disparities between visual and label information with modal-blended indiscriminate attention, replacing vanilla self-attention. Several combination blocks consisting of these two layers are integrated and embedded within the encoder-decoder structure to facilitate the exploration of meticulous visual label dependencies. Extensive experiments on two widely-used benchmarks, including MS-COCO and PASCAL VOC 2007, consistently demonstrate that PCMT could provide state-of-the-art results.