arXiv subjects

Yi-Chen Lu

Publications and source records attributed to Yi-Chen Lu.

3 recordsLinked to original sources

arXiv2026

Entanglement and firewalls in quantum circuit model of black hole evaporation

We reexamine the quantum circuit model of black hole evaporation proposed in a previous work (Class. Quantum Grav. 35, 235013, 2018) [1]. This tripartite model incorporates the following systems: black hole ($\mathbf{BH}$), just radiation ($\mathbf{JR}$), and early radiation ($\mathbf{ER}$). We apply a scrambling unitary matrix with a single parameter $\theta$ to the ground state of the qubits in infalling matter toward a black hole in order to generate initial qubit states of the black hole that are more general than those in [1]. Specifically, the scrambling unitary matrix reduces to no scrambling and maximum scrambling when $\theta=0$ and $\theta=\pi/2$, respectively. Our aim is to explore the role of quantum monogamy in the firewall formation between the black hole and radiation. In this model, entanglement and firewall formation depend on the black hole mass $M$ and the frequency of Hawking radiation $\omega$. For the initial state with $\theta=\pi/2$, a firewall emerges at an earlier stage of the evolution than with $\theta=0$. We also find that a firewall structure emerges between $\mathbf{BH}$ and $\mathbf{JR}$, and that the information is carried away by radiation for all values of $M\omega$, provided that $\theta$ lies within a certain analytically determined range. Following unitary gate dynamics, the initial black hole qubit state can be retrieved from its imprint on the final radiation state, which was originally hidden behind the black hole's horizon. These results may provide insight into the properties of multipartite entanglement due to the different initial states in the evolution of a quantum circuit model for black hole evaporation.

gr-qc↗

arXiv2025

Multimodal Chip Physical Design Engineer Assistant

Modern chip physical design relies heavily on Electronic Design Automation (EDA) tools, which often struggle to provide interpretable feedback or actionable guidance for improving routing congestion. In this work, we introduce a Multimodal Large Language Model Assistant (MLLMA) that bridges this gap by not only predicting congestion but also delivering human-interpretable design suggestions. Our method combines automated feature generation through MLLM-guided genetic prompting with an interpretable preference learning framework that models congestion-relevant tradeoffs across visual, tabular, and textual inputs. We compile these insights into a "Design Suggestion Deck" that surfaces the most influential layout features and proposes targeted optimizations. Experiments on the CircuitNet benchmark demonstrate that our approach outperforms existing models on both accuracy and explainability. Additionally, our design suggestion guidance case study and qualitative analyses confirm that the learned preferences align with real-world design principles and are actionable for engineers. This work highlights the potential of MLLMs as interactive assistants for interpretable and context-aware physical design optimization.

cs.AR↗

arXiv2025

Learning Library Cell Representations in Vector Space

We propose Lib2Vec, a novel self-supervised framework to efficiently learn meaningful vector representations of library cells, enabling ML models to capture essential cell semantics. The framework comprises three key components: (1) an automated method for generating regularity tests to quantitatively evaluate how well cell representations reflect inter-cell relationships; (2) a self-supervised learning scheme that systematically extracts training data from Liberty files, removing the need for costly labeling; and (3) an attention-based model architecture that accommodates various pin counts and enables the creation of property-specific cell and arc embeddings. Experimental results demonstrate that Lib2Vec effectively captures functional and electrical similarities. Moreover, linear algebraic operations on cell vectors reveal meaningful relationships, such as vector(BUF) - vector(INV) + vector(NAND) ~ vector(AND), showcasing the framework's nuanced representation capabilities. Lib2Vec also enhances downstream circuit learning applications, especially when labeled data is scarce.

cs.LG↗