upload dataset

We randomly sampled approximately 5,000 image pairs (336 × 336 resolution) from the GOT-10k dataset to fine-tune a Multimodal Large Language Model (MLLM) using reinforcement learning methods like GRPO, enabling it to excel in single-object tracking tasks.
Our recommended training framework is Easy-R1. [https://github.com/hiyouga/EasyR1]
R1-Track repository: [https://github.com/Wangbiao2/R1-Track].

train.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2ce543856cdc667c70e1d358216dcbee8b91edc368dbd556fd02735c0be9cc45
+size 251560710

validation.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1390b64ffd283cefb59213ec292f8fd185348601ca195392a9bc51cd47829984
+size 6521177