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..
Uncertainty-quantified Rollout Policy Adaptation for Unlabelled Cross-domain Video Temporal Grounding
Authors: Jian Hu, Zixu Cheng, Shaogang Gong, Isabel Guan, Jianye Hao, Jun Wang, Kun Shao
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | To evaluate the effectiveness of the proposed method, we conduct experiments on three widely used temporal grounding datasets: TACo S [38], Activity Net Captions [2], and Charades-STA [40]. Experiments on three datasets across six cross-domain settings show that URPA generalises well using only a few unlabelled target videos. |
| Researcher Affiliation | Collaboration | Jian Hu1, Zixu Cheng1 Shaogang Gong1, Isabel Guan2, Jianye Hao3 , Jun Wang4 , Kun Shao3 1Queen Mary University of London, 2Hong Kong University of Science and Technology, 3Huawei Noah s Ark Lab, 4University College London |
| Pseudocode | No | No pseudocode or algorithm block is explicitly presented in the paper. The methodology is described in natural language and mathematical equations. |
| Open Source Code | Yes | Our codes can be found in the supplemental material , and will be released upon publication. Codes are given in supplemental materials. |
| Open Datasets | Yes | To evaluate the effectiveness of the proposed method, we conduct experiments on three widely used temporal grounding datasets: TACo S [38], Activity Net Captions [2], and Charades-STA [40]. |
| Dataset Splits | Yes | Activity Net Captions contains approximately 20000 untrimmed You Tube videos annotated with 100000 natural language descriptions. Following the standard split, we use 37417 sentence-video pairs for training, and 17031 for testing. TACo S consists of 127 cooking-related videos. We adopt the public split, which includes 10146 and 4083 query-segment pairs for training and testing, respectively. Charades-STA is built on the Charades dataset, containing 12408 training and 3720 testing moment-query pairs. |
| Hardware Specification | Yes | All models are implemented in Py Torch on 32 NVIDIA V100 GPUs. |
| Software Dependencies | No | All models are implemented in Py Torch on 32 NVIDIA V100 GPUs. No specific version numbers for PyTorch or other software dependencies are provided. |
| Experiment Setup | Yes | In all experiments, the maximum prompt length is set to 4096, the maximum response length to 2048, the number of rollouts to 8, batch size to 16, and we train for 1 epoch. |