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..
NegoCollab: A Common Representation Negotiation Approach for Heterogeneous Collaborative Perception
Authors: CONGZHANG SHAO, Quan Yuan, Guiyang Luo, Yue Hu, Danni Wang, Liu Yilin, Rui Pan, Bo Chen, Jinglin Li
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | The experimental results demonstrate that Nego Collab significantly outperforms existing methods in common representation-based collaboration approaches. 4 Experiment 4.2 Quantitative Analysis 4.3 Ablation Study |
| Researcher Affiliation | Academia | 1State Key Laboratory of Networking and Switching Technology, Beijing University of Posts and Telecommunications 2Cooperative Medianet Innovation Center, Shanghai Jiaotong University EMAIL |
| Pseudocode | No | The paper describes the framework, sender, receiver, and training process with detailed explanations and supporting figures (Figure 2, Figure 5, Figure 6, Figure 7, Figure 8) but does not include any explicitly labeled 'Pseudocode' or 'Algorithm' block. |
| Open Source Code | Yes | 5. Open access to data and code Question: Does the paper provide open access to the data and code, with sufficient instructions to faithfully reproduce the main experimental results, as described in supplemental material? Answer: [Yes] Justification: The code has been open-sourced on github. |
| Open Datasets | Yes | We evaluated each method on the OPV2V-H Lu et al. (2024), V2V4Real Xu et al. (2023c), and DAIR-V2X Yu et al. (2022) datasets, as shown in Table 1 and Table 2. A.1 Dataset OPV2V-H. OPV2V-H Lu et al. (2024) dataset contains 73 scenes... DAIR-V2X. DAIR-V2X Yu et al. (2022) is a real-world collaborative perception dataset. V2V4Real. V2V4Real Xu et al. (2023c) is a real-world Vehicle-to-Vehicle (V2V) cooperative perception dataset. |
| Dataset Splits | No | The paper refers to using OPV2V-H, V2V4Real, and DAIR-V2X datasets, which are standard benchmarks, and describes training stages. However, it does not explicitly provide specific percentages or counts for training, validation, or test splits within the paper's text. The NeurIPS checklist states that 'the division of the optimizer and dataset can be viewed in the code,' implying these details are not in the paper itself. |
| Hardware Specification | Yes | A.2 Training Setup We conducted testing and training using a single RTX 4090 GPU, with an initial learning rate of 0.001 and Adam optimizer for parameter adjustment. The first training phase required approximately 4-12 GPU hours with about 23GB memory usage, while the second phase took around 2-5 GPU hours consuming approximately 14GB memory. |
| Software Dependencies | No | The paper mentions using an 'Adam optimizer' but does not specify any software libraries or frameworks with their version numbers (e.g., Python, PyTorch, TensorFlow, CUDA versions) that would be needed for replication. |
| Experiment Setup | Yes | A.2 Training Setup We conducted testing and training using a single RTX 4090 GPU, with an initial learning rate of 0.001 and Adam optimizer for parameter adjustment. The training process consists of two stages: The objective of the first stage is to negotiate common representations and to enable the sender-receiver to transform features from the local representation to and from the common representation. The training loss includes two components: cyclic distribution consistency loss and multi-dimensional alignment loss. The objective of the second stage is to adapt the framework to downstream collaborative tasks. |