SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2606.06819 · ICML · 2026

VideoSEG-O3: A Multi-turn Reinforcement Learning Framework for Reasoning Video Object Segmentation

Jingdong Wang, Sen Yang, Wankou Yang, Jiedong Zhuang, Ming Dai, Boyuan Tong, Boqiang Duan, Keyframe Mask

arXiv · PDF · Open in the Atlas

Code that ran

We lifted 12 functions out of this paper's own repositories and ran 12 of them in a sandbox. "Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results.

RepositoryRoleRan
Dmmm1997/VideoSEG-O3 canonical 12 of 12
FunctionStatusWhere it lives
dice_loss Ran Dmmm1997/VideoSEG-O3/projects/open_r1/trainer/sam_loss.py
code served (permissive licence) · get_code("27e0ec37f99eb303")
format_reward Ran Dmmm1997/VideoSEG-O3/projects/open_r1/rewards.py
code served (permissive licence) · get_code("09e6ed2ece2264f7")
get_mask_from_json Ran Dmmm1997/VideoSEG-O3/projects/open_r1/dataset.py
code served (permissive licence) · get_code("555397b8c8e37d9b")
iou_loss Ran Dmmm1997/VideoSEG-O3/projects/open_r1/trainer/sam_loss.py
code served (permissive licence) · get_code("e40a43378d8a1fb1")
lr_keyframe_vs_mask_reward_dense Ran Dmmm1997/VideoSEG-O3/projects/open_r1/rewards_dense.py
code served (permissive licence) · get_code("f00c0b416be367d1")
lr_progressive_keyframe_reward_dense Ran Dmmm1997/VideoSEG-O3/projects/open_r1/rewards_dense.py
code served (permissive licence) · get_code("8e18f662aafb6741")
lr_vtg_iou_reward_dense Ran Dmmm1997/VideoSEG-O3/projects/open_r1/rewards_dense.py
code served (permissive licence) · get_code("ce95cb2555ca57ce")
resize_longest Ran Dmmm1997/VideoSEG-O3/projects/open_r1/dataset.py
code served (permissive licence) · get_code("f198b4c1897d05e3")
resize_shortest Ran Dmmm1997/VideoSEG-O3/projects/open_r1/dataset.py
code served (permissive licence) · get_code("a95077549e0188f3")
sigmoid_bce_loss Ran Dmmm1997/VideoSEG-O3/projects/open_r1/trainer/sam_loss.py
code served (permissive licence) · get_code("96a68d44c9a31965")
think_format_reward Ran Dmmm1997/VideoSEG-O3/projects/open_r1/rewards.py
code served (permissive licence) · get_code("3bfb2a61622e10cb")
think_select_format_reward Ran Dmmm1997/VideoSEG-O3/projects/open_r1/rewards.py
code served (permissive licence) · get_code("fd576996738fc5b8")

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

Reasoning Video Object Segmentation (RVOS) demands a sophisticated integration of temporal dynamics, spatial details, and linguistic reasoning to achieve precise pixel-level localization. Existing methods are limited to reasoning over fixed initial inputs and lack the capacity to actively acquire further visual evidence, which is often essential for resolving complex references in long or intricate videos. To address this, we propose VideoSEG-O3, the first multi-turn reinforcement learning framework for RVOS that emulates the human "coarse-to-fine" cognitive process. It employs a multi-turn temporal-spatial chain-of-thought to capture fine-grained details by iteratively pinpointing critical intervals and keyframes. Additionally, to enable the policy to perceive segmentation quality beyond mere text probability of [SEG] during the RL stage, we introduce SEG-aware logit calibration, which integrates pixel-wise segmentation feedback directly into the token-level logits. Furthermore, we design a decoupled thinking trace to hierarchically decompose the reasoning process into temporal, spatial, and linguistic dimensions, and construct VTS-CoT, a specialized cold-start dataset featuring comprehensive reasoning trajectories. The code and models will be released at https:// github.com/Dmmm1997/VideoSEG-O3.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2606.06819")
get_code_for_paper("2606.06819")
have("2606.06819")

Connect an agent — have() is free.