{ "schema_version": "paper_public_manifest_v1", "paper_id": "mocap_2026_10", "slug": "mocap", "title": "MoCaP: Human Video-Driven Zero-Shot Mobile Manipulation", "authors": [], "source": { "arxiv_id": "2610.20998", "pdf_url": "https://arxiv.org/pdf/2610.20998", "project_url": "", "github_url": "", "huggingface_url": "", "original_source": "https://arxiv.org/abs/2610.20998" }, "site": { "post_url": "/posts/mocap", "canonical_url": "https://haiguangboy.com/posts/mocap", "cover_image": "https://static.haiguangboy.com/papers/mocap/cover.webp" }, "taxonomy": { "domain": "embodied_ai", "track": "action_generation", "tasks": [ "embodied_ai", "action_generation", "robotics", "state_prediction", "Embodied Intelligence", "Mobile Manipulation", "Robot Navigation", "Human Video", "mocap" ], "related_topics": [ { "paper_id": "an_open_foundation_model_towards_2026_07", "title": "An_Open_Foundation_Model_Towards", "url": "https://haiguangboy.com/posts/an_open_foundation_model_towards", "relation": "contrast", "summary": "All learning components frozen, no task-specific training contradicts joint co-training on heterogeneous human-robot data being a structurally suboptimal approach", "strength": "strong" }, { "paper_id": "omega_0_a_latent_predictive_world_action_model_for_concurrent_humanoid_loco_mani_2026_08", "title": "The more accurate the future prediction, the more sluggish the robot becomes", "url": "https://haiguangboy.com/posts/omega-0", "relation": "contrast", "summary": "Re-observe after the base moves, then plan the full arm contradicts problem framing: existing VLAs are either arm-centric or treat the base and arm as two uncoordinated subsystems", "strength": "strong" }, { "paper_id": "learning_a_thousand_tasks_in_a_day_2026_08", "title": "1000 tasks in 1 day, thanks to inductive bias", "url": "https://haiguangboy.com/posts/mt3-thousand-tasks", "relation": "same_track", "summary": "Zero-shot still relies on calibration, depth, and off-the-shelf execution modules validates perception dependency: vision only, single camera, no touch, relies on accurate segmentation", "strength": "strong" }, { "paper_id": "dextouch_wm_learning_action_conditioned_tactile_world_models_from_human_touch_fo_2026_09", "title": "DexTouch-WM: Human Tactile Pretraining", "url": "https://haiguangboy.com/posts/dextouch-wm", "relation": "same_track", "summary": "Zero-shot still relies on calibration, depth, and off-the-shelf execution modules validates that human data is not a direct transfer without hardware or robot data", "strength": "strong" }, { "paper_id": "google_blog_gemini_robotics_er_2_20260730_2026_07", "title": "Introducing Gemini Robotics ER 2", "url": "https://haiguangboy.com/posts/google_blog_gemini_robotics_er_2_20260730", "relation": "same_track", "summary": "Code agents use execution feedback to coordinate navigation and grasping validates the direct implication for navigation practitioners: navigation is not absorbed by the foundation model, but demoted to a tool invoked by the foundation model", "strength": "strong" }, { "paper_id": "mint_a_unified_model_for_world_space_camera_and_hand_motion_estimation_from_scal_2026_09", "title": "MINT_A_Unified_Model_for_World_Space_Camera_and_Hand_Motion_Estimation_from_Scal", "url": "https://haiguangboy.com/posts/mint_a_unified_model_for_world_space_camera_and_hand_motion_estimation_from_scal", "relation": "same_track", "summary": "VGGT relative pose obtains scale through robot depth validates the judgment behind two-stage training: learning a stable hand-camera coupled geometric relationship and correcting the absolute scale of the camera trajectory are two subproblems that can be solved separately—the former requires large amounts of diverse but noisy data, while the latter requires only a small amount of high-precision data", "strength": "strong" } ] }, "analyst_take": { "type": "author_opinion", "text": "💡 Engineering Insights\nThe use of human video can be extended to memory and approach experience directly invoked at execution time. This echoes the actuator activation conditions discussed in \"Towards Agentic Robots\": choosing the right skill also requires bringing the robot into a state where the skill can be executed.\nMoCaP demonstrates an experience-reuse path that does not update model weights. Whether it can scale still depends on whether the full manipulation success rate benefits stably after changing environments, changing recordings, and unifying the backend." }, "ruling": { "importance_score": 3.0, "one_sentence": "MoCaP: Human Video-Driven Zero-Shot Mobile Manipulation" }, "asset_base_url": "https://static.haiguangboy.com/papers/mocap", "assets": [ { "type": "cover_image", "object_key": "papers/mocap/cover.webp", "url": "https://static.haiguangboy.com/papers/mocap/cover.webp", "content_type": "image/webp", "upload_status": "pending", "role": "post_cover" }, { "type": "pdf_screenshot", "object_key": "papers/mocap/page_01.webp", "url": "https://static.haiguangboy.com/papers/mocap/page_01.webp", "content_type": "image/webp", "upload_status": "pending", "role": "paper_first_page" }, { "type": "pdf_screenshot", "object_key": "papers/mocap/key_figure.webp", "url": "https://static.haiguangboy.com/papers/mocap/key_figure.webp", "content_type": "image/webp", "upload_status": "pending", "role": "method_figure" }, { "type": "public_manifest", "object_key": "papers/mocap/public_manifest.json", "url": "https://static.haiguangboy.com/papers/mocap/public_manifest.json", "content_type": "application/json; charset=utf-8", "upload_status": "pending", "role": "public_manifest" }, { "type": "public_brief", "object_key": "papers/mocap/public_brief.md", "url": "https://static.haiguangboy.com/papers/mocap/public_brief.md", "content_type": "text/markdown; charset=utf-8", "upload_status": "pending", "role": "public_brief" } ], "published_at": "2026-10-18T17:34:32+08:00", "created_at": "2026-10-18T17:34:32+08:00", "updated_at": "2026-10-18T17:34:32+08:00" }