Intrinsic Sequence-Likelihood Confidence in Retrieval-Dominated Extractive QA: Two Pre-Specified Negatives, and What They Do and Do Not Attribute
研究探讨了在以检索为主的抽取式问答中，使用预设负例的方法来解决模型自信度信号不足的问题，但实验表明该方法未能显著提升性能。
0 citationsRead paper
Selected work
Representative Papers
PennyLane-Lightning MPI: A massively scalable quantum circuit simulator based on distributed computing in CPU clusters
Quantum circuit simulation faces severe computational bottlenecks due to the exponential growth of Hilbert space dimension with qubit count. To address this, we propose an index-dependent, gate-specific parallelization strategy that exploits the locality of single-qubit gates and the block structure of the state vector, enabling efficient distributed-memory quantum state evolution on CPU clusters. Our approach employs MPI for inter-node parallelism, integrates with the PennyLane plugin interface, and leverages fine-grained vector partitioning to maximize scalability. Experiments on standard multi-core CPU clusters successfully simulate 41-qubit circuits using over 100,000 concurrent processes, outperforming generic unitary-matrix-based methods in both speed and memory efficiency. The implementation achieves strong scaling and has been deployed as the high-performance simulation backend of a national cloud-based quantum computing platform in Korea.
0 citationsRead paper
SRVP: Strong Recollection Video Prediction Model Using Attention-Based Spatiotemporal Correlation Fusion
In video prediction, RNN-based models suffer from progressive degradation of appearance details due to long-term memory accumulation, leading to significant quality deterioration as prediction horizon increases. To address this, we propose the Strongly Retrospective Video Prediction (SRVP) model, which introduces a novel dual-attention mechanism—integrating Standard Attention (SA) and Reinforced Feature Attention (RFA)—to explicitly decouple and jointly model spatiotemporal dependencies, thereby overcoming RNNs’ inherent detail-forgetting bottleneck. SRVP employs differentiable scaled dot-product attention for spatiotemporal feature fusion and enables end-to-end learning of high-fidelity spatiotemporal representations. Evaluated on three standard benchmarks, SRVP substantially mitigates quality degradation, achieving average improvements of +2.1 dB in PSNR and +0.035 in SSIM over strong RNN baselines, while matching the prediction accuracy of state-of-the-art RNN-free approaches.
0 citationsRead paper
Recent publications
Latest Papers
Intrinsic Sequence-Likelihood Confidence in Retrieval-Dominated Extractive QA: Two Pre-Specified Negatives, and What They Do and Do Not Attribute
研究探讨了在以检索为主的抽取式问答中，使用预设负例的方法来解决模型自信度信号不足的问题，但实验表明该方法未能显著提升性能。
0 citationsRead paper
PennyLane-Lightning MPI: A massively scalable quantum circuit simulator based on distributed computing in CPU clusters
Quantum circuit simulation faces severe computational bottlenecks due to the exponential growth of Hilbert space dimension with qubit count. To address this, we propose an index-dependent, gate-specific parallelization strategy that exploits the locality of single-qubit gates and the block structure of the state vector, enabling efficient distributed-memory quantum state evolution on CPU clusters. Our approach employs MPI for inter-node parallelism, integrates with the PennyLane plugin interface, and leverages fine-grained vector partitioning to maximize scalability. Experiments on standard multi-core CPU clusters successfully simulate 41-qubit circuits using over 100,000 concurrent processes, outperforming generic unitary-matrix-based methods in both speed and memory efficiency. The implementation achieves strong scaling and has been deployed as the high-performance simulation backend of a national cloud-based quantum computing platform in Korea.
0 citationsRead paper
SRVP: Strong Recollection Video Prediction Model Using Attention-Based Spatiotemporal Correlation Fusion
In video prediction, RNN-based models suffer from progressive degradation of appearance details due to long-term memory accumulation, leading to significant quality deterioration as prediction horizon increases. To address this, we propose the Strongly Retrospective Video Prediction (SRVP) model, which introduces a novel dual-attention mechanism—integrating Standard Attention (SA) and Reinforced Feature Attention (RFA)—to explicitly decouple and jointly model spatiotemporal dependencies, thereby overcoming RNNs’ inherent detail-forgetting bottleneck. SRVP employs differentiable scaled dot-product attention for spatiotemporal feature fusion and enables end-to-end learning of high-fidelity spatiotemporal representations. Evaluated on three standard benchmarks, SRVP substantially mitigates quality degradation, achieving average improvements of +2.1 dB in PSNR and +0.035 in SSIM over strong RNN baselines, while matching the prediction accuracy of state-of-the-art RNN-free approaches.
0 citationsRead paper