Add files using upload-large-folder tool

.gitattributes

@@ -58,3 +58,6 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 # Video files - compressed
 *.mp4 filter=lfs diff=lfs merge=lfs -text
 *.webm filter=lfs diff=lfs merge=lfs -text
+train/data_sft.jsonl filter=lfs diff=lfs merge=lfs -text
+train/data_rl.jsonl filter=lfs diff=lfs merge=lfs -text
+train/annotations.jsonl filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,80 @@
+---
+license: cc-by-nc-4.0
+task_categories:
+- video-text-to-text
+language:
+- en
+tags:
+- social-reasoning
+- gaze
+- gesture
+- video-question-answering
+size_categories:
+- 100K<n<1M
+configs:
+- config_name: default
+  data_files:
+  - split: train
+    path: train/annotations.jsonl
+  - split: test
+    path: test/annotations.jsonl
+---
+
+# GRASP: Learning to Ground Social Reasoning in Multi-Person Non-Verbal Interactions
+
+[📜 Paper](https://arxiv.org/abs/2605.15764) · [🌐 Project Page](https://social-reaoning.github.io/grasp/) · [💻 Code](https://github.com/Social-Reaoning/grasp) · [🤖 Model](https://huggingface.co/interlive/GRASP-Qwen3-VL-8B)
+
+GRASP is a large-scale social reasoning dataset that connects high-level social QA with fine-grained gaze and deictic gesture events, organized by a 16-category taxonomy spanning gaze (T1–T6), gesture (G1–G6), and joint gaze–gesture reasoning (J1–J4), together with **GRASP-Bench** for evaluation.
+
+## Structure
+
+```
+train/
+├── annotations.jsonl        # all training QA, one row per QA (shown in the Dataset Viewer)
+├── data_sft.jsonl           # open-ended QA with structured reasoning traces (SFT, training format)
+├── data_rl.jsonl            # MCQ with ground-truth social events (SGR / GRPO, training format)
+└── videos/                  # per-QA video clips, tar parts (<=9GB each)
+    ├── avsbench_part001.tar         # extracts to avsbench/video/*.mp4
+    ├── embody3d_part001.tar ...
+    └── ...
+test/
+├── annotations.jsonl        # GRASP-Bench QA, one row per QA (shown in the Dataset Viewer)
+└── grasp_bench.tar          # GRASP-Bench: json/*.json + video/*.mp4 (1,196 items)
+```
+
+`annotations.jsonl` columns: `id`, `video` (path inside the extracted archives), `source`, `category`, `difficulty`, `format` (`mcq` / `open_ended`), `question`, `options`, `answer`, `response` (structured reasoning target for open-ended training QA).
+
+Each training JSONL line is a list of input specs:
+
+```json
+[{"type": "video", "path": "avsbench/video/<clip>.mp4", "fps": 2.0, ...},
+ {"type": "text", "content": "<question>", "output": false},
+ {"type": "text", "content": "<answer>", "output": true}]
+```
+
+Video `path` fields are relative to the extracted `train/videos/` root. Extract all tars into one directory and point training at it:
+
+```bash
+for t in train/videos/*.tar; do tar -xf "$t" -C /path/to/video_root; done
+```
+
+GRASP-Bench items are one JSON per QA with `clip_file`, `qa.category` (T1–T6 / G1–G6 / C1–C4 for joint), `qa.question`, `qa.options`, `qa.answer`.
+
+## Usage
+
+Training and evaluation code: https://github.com/Social-Reaoning/grasp
+
+## License
+
+GRASP annotations (QA pairs and social event labels) are released under [CC BY-NC 4.0](https://creativecommons.org/licenses/by-nc/4.0/) for non-commercial research use only. Video clips are derived from AVSBench, Embody3D, FriendsMMC, Social-IQ, SocialGesture, TVQA, and Werewolf Among Us, and remain subject to the licenses and terms of use of their original sources. Commercial use is not permitted.
+
+## Citation
+
+```bibtex
+@article{kim2026grasp,
+  title={GRASP: Learning to Ground Social Reasoning in Multi-Person Non-Verbal Interactions},
+  author={Kim, Junho and Cao, Xu and Yang, Houze and Boote, Bikram and Jojic, Ana and Ryan, Fiona and Lai, Bolin and Lee, Sangmin and Rehg, James M},
+  journal={arXiv preprint arXiv:2605.15764},
+  year={2026}
+}
+```

test/grasp_bench.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:877ef85d5ad3ba6a8e3a580607c77495814056ceb53b6e81184f85365b386dec
+size 419584000

train/annotations.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0735e8deb3f846f5c4becf9db12b094b6290b05b8b166c2d1077dcc390081de9
+size 195880264

train/data_rl.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b246e0150f6b5065d78cb401166cf35fafe8a3f80330d426be07394dffd49deb
+size 238099900

train/data_sft.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ab1ea82d92ef563b6da995983f6fbdb416416ecfb597ccfb3cc2235919adacfa
+size 94361874

train/videos/avsbench_part001.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:602680e42402f3506d855be6d83ed49c7fbd6ab099a3fd0b8b73b676e4b79f5c
+size 2686556160

train/videos/embody3d_part001.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d67fc981051d512fadb0d38d699a591fda88e76d7898260875b60a9af94248bd
+size 9669027840

train/videos/embody3d_part002.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:86953458c8aa53e3c9d80a5204ed66afd1ebe3eecd1b5ad687cc74b91b763b34
+size 9671802880

train/videos/embody3d_part003.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c1825d6342ccb247456a0a96b1be5b32ad72553828bfb8ffb57fb053c7817279
+size 9666580480

train/videos/embody3d_part004.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:29f8b98a9d73a493a5321b926a34d0fd7a2ad53d5906cbf642fbf676440597ed
+size 9675653120

train/videos/embody3d_part005.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1b6bdea3aed0736e549c2ee126cbdac7111fb603013b7a837cc540b7e4361c56
+size 9674803200

train/videos/embody3d_part006.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7777f5a4cffb8232bcf0b5b7be931bc6440235cb3fead9351e13ca996e455918
+size 6075412480

train/videos/friendsmmc_part001.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a2eb96ca3e1532c7366c86d9beb18d20450e06a5c9496815b12e4b4eaee71296
+size 9679503360

train/videos/friendsmmc_part002.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:65e6ade65d3cb7884b60074fe3559b264055f5a7b1904e9a66bb14df8b9e6453
+size 9681336320

train/videos/friendsmmc_part003.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8de330d95c4fb80447e57bc0467980bfed000254a4e2340b0afba2f86b67317d
+size 9681428480

train/videos/friendsmmc_part004.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2d6bec5e35f9c7fdf2c3396633cd66162c16bf822fb08067ceeaef9d95303003
+size 6650265600

train/videos/social-iq_part001.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fb857460e415fc7751231372301ee47fdc1052c7371ec23c2c09d61426839516
+size 2633584640

train/videos/social_gesture_part001.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:74ba5b471de3055caf7ce918d40c875b44218db7586b09c3d100c3c0380fbb0e
+size 5314713600

train/videos/tvqa_part001.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bee582226f15c0996d6430dcced3c85ecb27f56d72af9cefa2b1cecec881704d
+size 9718804480

train/videos/tvqa_part002.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:56e865c997bcf48b2ac530e9302dd87008a43d7b95aad313538f9c0199684835
+size 9736591360

train/videos/tvqa_part003.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3661895fa56db271d30dd3b0211f3c62219fe05bdb4b4d010e60f95e95cc4fd0
+size 9705052160

train/videos/tvqa_part004.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5e784762303eb214c8ba83ec9b49985f017d5bfcb08680e58a4d87e71d4660ac
+size 9716776960

train/videos/tvqa_part005.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2e35c6ac5d3e0d207999f788720309935393ab2a47539c13987de8fcd91d3927
+size 5632409600

train/videos/werewolf_part001.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:16783607d58f96d2e8006b3646262495d2bdda214fe2be880a2f4df7a4093031
+size 9700372480

train/videos/werewolf_part002.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1c89a24e45e7222b79af02e8479a7251bdd2b38b97087aee1787a4d61c6facfd
+size 5052794880