Large language models (LLMs) have recently emerged as a promising approach for automating Verilog code generation; however, existing methods primarily emphasize syntactic correctness and often rely on commercial models or external verification tools, which introduces concerns regarding cost, data privacy, and limited guarantees of functional correctness. This work proposes a unified multi-agent framework for reasoning-oriented training data generation with integrated testbench-driven verification, enabling locally fine-tuned LLMs, SiliconMind-V1, to iteratively generate, test, and debug Register-Transfer Level (RTL) designs through test-time scaling. Experimental results on representative benchmarks (VerilogEval-v2, RTLLM-v2, and CVDP) demonstrate that the proposed approach outperforms the state-of-the-art QiMeng-CodeV-R1 in functional correctness while using fewer training resources.

Recent advancements in large vision-language models (LVLM) have significantly enhanced their ability to comprehend visual inputs alongside natural language. How-ever, a major challenge in their real-world application is hallucination, where LVLMs generate non-existent visual elements, eroding user trust. The underlying mechanism driving this multimodal hallucination is poorly understood. Minimal research has illuminated whether contexts such as sky, tree, or grass field involve the LVLM in hallucinating a frisbee. We hypothesize that hiddenfactors, such as objects, contexts, and semantic foreground-background structures, induce hallucination. This study proposes a novel causal approach: a hallucination probing system to identify these hidden factors. By analyzing the causality between images, text prompts, and network saliency, we systematically ex-plore interventions to block these factors. Our experimen-tal findings show that a straightforward technique based on our analysis can significantly reduce hallucinations. Additionally, our analyses indicate the potential to edit network internals to minimize hallucinated outputs.

Learning from noisy-labeled data is crucial for real-world applications. Traditional Noisy-Label Learning (NLL) methods categorize training data into clean and noisy sets based on the loss distribution of training samples. However, they often neglect that clean samples, especially those with intricate visual patterns, may also yield substantial losses. This oversight is particularly significant in datasets with Instance-Dependent Noise (IDN), where mislabeling probabilities correlate with visual appearance. Our approach explicitly distinguishes between clean $v s$. noisy and easy $v s$. hard samples. We identify training samples with small losses, assuming they have simple patterns and correct labels. Utilizing these easy samples, we hallucinate multiple anchors to select hard samples for label correction. Corrected hard samples, along with the easy samples, are used as labeled data in subsequent semi-supervised training. Experiments on synthetic and real-world IDN datasets demonstrate the superior performance of our method over other state-of-the-art NLL methods.

Privacy-preserving neural inference helps protect both the user input data and the model weights from being leaked to others during the inference of a deep learning model. To achieve data protection, the inference is often performed within a secure domain, and the final result is revealed in plaintext. Nevertheless, performing the computations in the secure domain incurs about a thousandfold overhead compared with the insecure version, especially when the involved operations of the entire model are mapped to the secure domain, which is the computation scheme adopted by the existing works. This work is inspired by the transfer learning technique, where the weights of some parts of the model layers are transferred from a publicly available, pre-built deep learning model, and it opens a door to further boost the execution efficiency by allowing us to do the secure computations selectively on parts of the transferred model. We have built a compiler framework, SecureTVM, to automatically translate a trained model into the secure version, where the model layers to be protected can be selectively configured by its model provider. As a result, SecureTVM outperforms the state of the art, CrypTFlow2, by a factor of 55 for the transfer learning model. We believe that this work takes a step forward toward the practical uses of privacy-preserving neural inference for real-world applications.