Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in Coakley et alK. L. Coakley, T. Snelleman, H. Hoos, and O. E. Gundersen, "The embrace of open science: An analysis of a decade of AI research and 56 800 conference papers," Under Review, 2026..

Event-Guided Consistent Video Enhancement with Modality-Adaptive Diffusion Pipeline

Authors: Kanghao Chen, Zixin Zhang, Guoqiang Liang, Lutao Jiang, Zeyu Wang, Yingcong Chen

NeurIPS 2025 | Venue PDF | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Our experiments demonstrate state-of-the-art performance across challenging scenarios (i.e., varying illumination) and sensor-based settings (e.g., event-only, RGB-only), highlighting the generalization of our framework.
Researcher Affiliation Academia Kanghao Chen AI Thrust, HKUST(GZ) EMAIL Zixin Zhang AI Thrust, HKUST(GZ) EMAIL Guoqiang Liang Nanyang Technological University Lutao Jiang AI Thrust, HKUST(GZ) Zeyu Wang CMA Thrust, HKUST(GZ) CSE, HKUST Ying-Cong Chen AI Thrust, HKUST(GZ) EMIA, HKUST EMAIL
Pseudocode No The paper describes the methodology in prose and architectural diagrams (e.g., Figure 2) but does not include any explicitly labeled 'Pseudocode' or 'Algorithm' blocks.
Open Source Code Yes We provide open access to data and code in the supplemental material
Open Datasets Yes To construct the evaluation dataset, we synthesize videos with varying illumination based on the SDE dataset [23] and SDSD dataset [45], referred to as VSDE and V-SDSD, respectively.
Dataset Splits No The paper mentions using 'SDE training dataset', 'SDSD training dataset', and 'SDE test set' for evaluation, but does not provide specific percentage splits or absolute sample counts for training, validation, and testing sets required to reproduce data partitioning.
Hardware Specification Yes The training process takes 2 days on 8 H100 GPUs.
Software Dependencies No The paper mentions fine-tuning 'pre-trained I2V diffusion models (i.e., Cog Video XI2V [51])' and employing the 'DDIM [40] sampler', but does not specify programming language versions or library versions (e.g., Python, PyTorch, CUDA versions).
Experiment Setup Yes The Lo RA rank is set to 128 for the image branch. For training, we use a learning rate of 1 10 4 and the Adam W optimizer. The model is trained for 30 epochs with gradient accumulation, resulting in an effective batch size of 64.