Ye Liu

Where are the evaluation results of other methods reported in the paper from?

1

Hi @yingtaoluo, may I ask that did you reproduce other methods (e.g. GeoSAN, LSTPM, etc.) on exactly the same datasets you used for STAN, or did you just copy the...

Unable to Reproduce 12 Epoch Results in the Paper

5

Hi @Jingkang50, thanks for your great work. I ran a 12-epoch training for PSGFormer by modifying the [config](https://github.com/Jingkang50/OpenPSG/blob/main/configs/psgformer/psgformer_r50_psg.py) from `40/60 epochs` to `8/12 epochs`, and setting `eval_pan_rels=False` according to your...

[Feature] Any plans to add support for video LLMs?

2

Some video LLMs are also based on LLaVA codebase, e.g., Video-ChatGPT.

Code for Dataset Generation

2

Hi @mmaaz60, thanks for sharing this great work! I was wondering whether you have a plan to share the code for semi-automated dataset generation (the pipeline of using Katna to...

[Bug] The video processor of InternVL-3.5-1B-HF is inconsistent with image processor

1

### Checklist - [x] 1. I have searched related issues but cannot get the expected help. - [x] 2. The bug has not been fixed in the latest version. -...

Hi @srama2512, thanks for sharing this project! I notice that in `data/dataset/nlq_aug_naq_official_v2/train.json`, every sentence ends with ".?" even if it is not a question at all. I was wondering whether...