AI Paper Atlas

Research map / Video / Video segmentation and tracking

GT-SVJ: Generative-Transformer-Based Self-Supervised Video Judge For Efficient Video Reward Modeling

Shivanshu Shekhar, Uttaran Bhattacharya, Raghavendra Addanki, Mehrab Tanjim, Somdeb Sarkhel, Tong Zhang

CVPR 2026 · Video › Video segmentation and tracking · approach: Self-supervised video representation learning · 2 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
pretrain a contrastive discriminative model before reward tuning (ablation)GT-SVJ reward model · GenAI-Bench (w/ties)preference accuracy (%): 47.62 -> 64.26Improved
pretrain a contrastive discriminative model before reward tuning (ablation)GT-SVJ reward model · MonteBench (w/o ties)preference accuracy (%): 73.59 -> 77.76Improved
add variable-length video truncation (p=0.25) (ablation)GT-SVJ discriminative model · GenAI-Bench (w/ties)preference accuracy (%): 49.36 -> 64.26Improved
add variable-length video truncation (p=0.25) (ablation)GT-SVJ discriminative model · MonteBench (w/ties)preference accuracy (%): 62.79 -> 66.36Improved
add perturbed videos as negatives alongside generated videos (ablation)GT-SVJ discriminative model · validation alignment accuracyaccuracy: 2-3% better validation accuracy, faster convergenceImproved
add perturbed videos as negatives alongside generated videos (ablation)GT-SVJ discriminative model · training dynamicsgradient norm / lossImproved
apply LoRA to middle third instead of last third of transformer layers (ablation)CogVideoX-2B reward model · validation loss and accuracyvalidation accuracyImproved
apply LoRA to last third instead of middle third (ablation)CogVideoX-2B reward model · training speedtraining time: about 1.5x faster training, marginal performance decreaseImproved
sample perturbations by inverse gradient-decay difficulty instead of uniformlyGT-SVJ discriminative model · negative samplingtraining dynamicsNo clear effect
use a generative video model backbone instead of a VLM as reward modelGT-SVJ · GenAI-Bench (w/ties)preference accuracy (%): 51.56 -> 64.26 vs VisionRewardImproved
use a generative video model backbone instead of a VLM as reward modelGT-SVJ · VideoReward-Bench (w/ties)preference accuracy (%): 61.26 -> 57.01 vs VideoRewardHurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy