Emergent Communication in Interactive Sketch Question Answering
Authors: Zixing Lei, Yiming Zhang, Yuxin Xiong, Siheng Chen
NeurIPS 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Our experimental results including human evaluation demonstrate that multi-round interactive mechanism facilitates targeted and efficient communication between intelligent agents with decent human interpretability. The code is available at here. |
| Researcher Affiliation | Collaboration | Zixing Lei1, Yiming Zhang2, Yuxin Xiong1, Siheng Chen1,3 chezacarss@sjtu.edu.cn; yz2926@cornell.edu; xyx1323@sjtu.edu.cn; sihengc@sjtu.edu.cn 1 Cooperative Medianet Innovation Center, Shanghai Jiao Tong University, 2 Cornell University, 3 Shanghai AI Laboratory |
| Pseudocode | No | The paper provides mathematical formulations and descriptions of the model, but does not include any explicitly labeled pseudocode or algorithm blocks. |
| Open Source Code | Yes | The code is available at here. |
| Open Datasets | Yes | Visual Question Answering (VQA) v2.0 [36] is a dataset that containas open-ended questions about images... Visual Genome (VG) [49] is a dataset that contains 108K images... |
| Dataset Splits | No | The paper describes training with a loss function and evaluating on VQA v2.0 and VG datasets but does not provide specific details on validation dataset splits (e.g., percentages, sample counts, or explicit validation set usage for hyperparameter tuning). |
| Hardware Specification | No | The paper does not provide any specific hardware details such as GPU models (e.g., NVIDIA A100), CPU types, or memory configurations used for experiments. |
| Software Dependencies | No | The paper references various software components and models (e.g., MCAN, Faster-RCNN, CLIP, ResNet) but does not specify their version numbers (e.g., 'PyTorch 1.9', 'TensorFlow 2.0'). |
| Experiment Setup | No | The paper describes the overall training objective (minimizing L) and the role of hyperparameter 'a', but it lacks specific details regarding concrete hyperparameter values such as learning rates, batch sizes, optimizer choices, or the number of training epochs. |