CoCoG: Controllable Visual Stimuli Generation Based on Human Concept Representations
Authors: Chen Wei, Jiachen Zou, Dietmar Heinke, Quanying Liu
IJCAI 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | The experiments with Co Co G indicate that 1) the reliable concept embeddings in Co Co G allows to predict human behavior with 64.07% accuracy in the THINGS-similarity dataset; 2) Co Co G can generate diverse objects through the control of concepts; 3) Co Co G can manipulate human similarity judgment behavior by intervening key concepts. |
| Researcher Affiliation | Academia | 1Southern University of Science and Technology, Shenzhen, China 2University of Birmingham, Birmingham, United Kingdom {weic3, zoujc2022}@mail.sustech.edu.cn, d.g.heinke@bham.ac.uk, liuqy@sustech.edu.cn |
| Pseudocode | No | The paper describes the method verbally and with diagrams, but does not include any structured pseudocode or algorithm blocks. |
| Open Source Code | Yes | The code of Co Co G is available at https://github.com/ncclab-sustech/Co Co G. |
| Open Datasets | Yes | We used the triplet odd-one-out similarity judgment task in the THINGS dataset [Hebart et al., 2023]. |
| Dataset Splits | No | The paper mentions using the THINGS Odd-one-out dataset to train and validate the concept encoder, but it does not provide specific details on the train/validation/test splits, such as percentages, sample counts, or explicit references to predefined splits in the main text. |
| Hardware Specification | No | The paper does not provide specific hardware details such as GPU models, CPU types, or memory specifications used for running the experiments. |
| Software Dependencies | No | The paper mentions using 'CLIP image encoder', 'pre-trained SDXL and IP-Adapter models', and 'U-Net', but it does not provide specific version numbers for any of these software dependencies. |
| Experiment Setup | No | The paper states 'Specific training parameters are shown in the Appendix,' indicating that these details are not provided in the main body of the paper. |