[
    {
        "id": "osp-19463",
        "type": "article-journal",
        "title": "LoGo: Local-Global Rewards for Consistent Long-Horizon Video Generation",
        "author": [
            {
                "family": "Ma",
                "given": "Ziqi"
            },
            {
                "family": "Sharma",
                "given": "Shreya"
            },
            {
                "family": "El Banani",
                "given": "Mohamed"
            },
            {
                "family": "Schwarz",
                "given": "Katja"
            },
            {
                "family": "Ye",
                "given": "Chongjie"
            },
            {
                "family": "Wu",
                "given": "Chao-Yuan"
            },
            {
                "family": "Fei-Fei",
                "given": "Li"
            },
            {
                "family": "Mildenhall",
                "given": "Ben"
            },
            {
                "family": "Gkioxari",
                "given": "Georgia"
            },
            {
                "family": "Johnson",
                "given": "Justin"
            },
            {
                "family": "Somepalli",
                "given": "Gowthami"
            }
        ],
        "URL": "https://omanscience.com/ar/articles/logo-local-global-rewards-for-consistent-long-horizon-video-generation",
        "language": "en",
        "issued": {
            "date-parts": [
                [
                    2026
                ]
            ]
        },
        "abstract": "Camera-controlled video models are rapidly advancing toward long generation horizons and complex camera control. A key failure mode is 3D inconsistency: as the camera moves, objects lose permanence and scene structures shift. Existing post-training techniques, which assign a single scalar reward to the entire generation, are poorly suited to correcting these inconsistencies over long horizons. We introduce LoGo, which blends global and spatially localized rewards for camera-controlled video models. The local reward provides fine-grained credit assignment, which substantially improves 3D consistency, while the global reward preserves camera following and video quality. Across three base models, LoGo shows a clear advantage on DL3DV and TrajectoryBench, a new benchmark for long-horizon, complex-camera-control generation that current evaluations lack. LoGo effectively reduces local object shifts, artifacts, and global scene changes, illustrating the importance of credit assignment in post-training video models. Project website: https://ziqi-ma.github.io/logo-website/"
    }
]