{"slug": "open-humanwam-50k-hours-of-egocentric-video-for-robot-wams", "title": "Open-HumanWAM: 50K Hours of Egocentric Video for Robot WAMs", "summary": "Peking University's DAGroup published Open-HumanWAM, an open recipe for pretraining world–action models on 50,000 hours of egocentric human video and post-training them into robot policies, with the arXiv paper (2611.06817), project site, Hugging Face models and dataset card going live in early November 2026. HumanWAM-Vanilla, after robot post-training, is reported at 92.08% success on RoboTwin 2.0 Clean2Clean and 68.88% on Clean2Random, while HumanWAM-Goal reaches 84.60% on the randomized split and averages 88.65% across Clean and Random. The release bundles the Open-HumanWAM-50K corpus with hand-motion and language annotation, open weights, training code and benchmark scripts, and reports an empirical log-linear relationship between pretraining hours and downstream robot success.", "body_md": "**Peking University's DAGroup published Open-HumanWAM**, an open recipe to pretrain **world–action models** on **fifty thousand hours** of egocentric human video and post-train them into robot policies that lead reported **RoboTwin 2.0** numbers among compared open checkpoints. The [arXiv paper](https://arxiv.org/abs/2611.06817?ref=explainx) ( **2611.06817** ), [project site](https://dagroup-pku.github.io/open-humanwam/?ref=explainx), [Hugging Face models](https://huggingface.co/collections/DAGroup-PKU/open-humanwam?ref=explainx), and [dataset card](https://huggingface.co/datasets/DAGroup-PKU/Open-HumanWAM-50K?ref=explainx) went live in early November 2026.\n\nIf you follow embodied AI because [human egocentric video is the scalable manipulation prior](https://explainx.ai/blog/egocentric-data-collection-robots-explained-2026), this release is one of the first to bundle **corpus, checkpoints, training code, and benchmark scripts** under a single branding umbrella rather than a leaderboard screenshot alone.\n\n## TL;DR\n\n| Question | Answer | \n|---|---|\n| Corpus | **Open-HumanWAM-50K** — curated egocentric video, hand motion + language | \n| Models | **HumanWAM-Vanilla** (human-only pretrain) and**HumanWAM-Goal** (human + robot, visual goals) | \n| Key sim result (post-trained) | Up to **~88.65%** average on**RoboTwin 2.0** Clean + Random per project tables | \n| Scaling claim | **Log-linear** link between pretraining hours and downstream success; diversity matters | \n| Open assets | [GitHub HumanNet branch](https://github.com/DAGroup-PKU/HumanNet/tree/open-humanscale?ref=explainx) , HF weights + dataset | \n| Builder angle | Reproduce scaling curves before betting your robot stack on proprietary human-video APIs | \n\n## What the paper adds beyond another VLA headline\n\nThe abstract on [Hugging Face Papers](https://huggingface.co/papers/2611.06817?ref=explainx) states the problem bluntly: raw human video scale does not automatically become generalizable robot policies without **curation** and a **training recipe**. Open-HumanWAM claims three contributions readers can verify if they download the assets:\n\n1. **A 50k-hour corpus** with dense hand-motion and natural-language annotation—not a token YouTube scrape without labels.\n2. **Systematic scaling experiments** in simulation that produce an empirical**log-linear law** between human-video hours and robot success after adaptation.\n3. **Two model lines** —HumanWAM-Vanilla (human pretrain only) and HumanWAM-Goal (joint human–robot with visual goals)—with**open weights** and evaluation harnesses.\n\nHumanWAM-Vanilla, after robot post-training, is reported at **92.08%** success on RoboTwin 2.0 **Clean2Clean** and **68.88%** on **Clean2Random**, with additional **RoboDojo** metrics in the paper text (about **14.17%** mean success and **19.00** mean progress score in the HF summary). HumanWAM-Goal pushes the harder randomized split to **84.60%**, averaging **88.65%** across Clean and Random in the post-trained regime described on the project site.\n\nThose numbers compete with other 2026 world–action releases we have covered—[Verne V0.5](https://explainx.ai/blog/verne-robotics-v0-5-open-weight-world-action-model-8b-libero-99-3-robotwin-81-gated-license-claim-check-2026) reported **81.0% / 73.4%** clean/random on RoboTwin with an 8B checkpoint, and [SLA continual pretraining](https://explainx.ai/blog/sla-spatial-language-action-model-continual-pretraining-vla-robotics-arxiv-2611-08373-2026) reported **72.3%** Clean2Random. **Do not treat leaderboard ranks as eternal**—each paper uses slightly different post-training budgets—but Open-HumanWAM's selling point is that you can **download their checkpoints and rerun the same RoboTwin scripts**.\n\n## HumanWAM-Vanilla: scaling human experience\n\nHumanWAM study protocol from egocentric pretrain through RoboTwin adaptation\n*Image: [PKU DAGroup](https://dagroup-pku.github.io/open-humanwam/?ref=explainx)*\n\nThe [project microsite](https://dagroup-pku.github.io/open-humanwam/?ref=explainx) walks through a three-step protocol:\n\n1. **Learn from people** — video, language, and hand motion in egocentric frames.\n2. **Adapt to robots** — post-train on**50 RoboTwin tasks** with**50 clean demonstrations** each.\n3. **Test generalization** — clean scenes, scene out-of-distribution, and randomized clutter.\n\nScaling charts compare **100 → 10,000 → 50,000 hours** under both **isolated attention** and **inverse-dynamics attention** variants. The authors emphasize gains **outside training scenes**, not only memorizing tabletop layouts.\n\nInverse-dynamics attention scaling from 100 to 50,000 hours of egocentric pretraining\n*Image: [PKU DAGroup](https://dagroup-pku.github.io/open-humanwam/?ref=explainx)*\n\nA second study splits **scaled-diversity** corpora (more tasks and scenes as hours grow) from **coverage-matched** corpora (more trajectories within fixed task/scene coverage). The site argues **diversity scaling** produces steeper transfer than repeating similar clips—a lesson for teams paying for [egocentric annotation APIs](https://explainx.ai/blog/perceptron-egocentric-1-0-api-9-dollars-video-hour-egohome-gpt-6-astra-one-eighth-cost-2026) without widening scene coverage.\n\nScaled-diversity versus coverage-matched corpus scaling curves\n*Image: [PKU DAGroup](https://dagroup-pku.github.io/open-humanwam/?ref=explainx)*\n\n## HumanWAM-Goal: visual goals and zero-shot behavior\n\nHumanWAM-Goal trains a **visual-goal-conditioned** world–action model on **hybrid human and robot data**. Before RoboTwin-specific post-training, the project lists **13.94%** average zero-shot success across Clean and Random, which they describe as leading among **successfully evaluated open-source zero-shot** checkpoints in their comparison table—not the same protocol as the post-trained **88.65%** figure.\n\nOn real hardware, the site reports **20.0% → 30.5%** average success across twenty tasks when moving from **robot-only** pretraining to hybrids with **15.9k** and **45.9k hours** of human data, with robot experience fixed at **14,618 hours** in that ablation.\n\nQualitative demos show **goal-conditioned in-context learning**: the policy keeps weights frozen while the visual target changes—useful for teams thinking about [Natura-1-style dexterous transfer](https://explainx.ai/blog/nature-will-natura-1-dexterous-hands-foundation-model-egocentric-human-video-2026) without collecting hundreds of new teleop demos per SKU.\n\nLong-horizon clips include **desk organization** (fourteen stages) and mobile tidying on **Astribot S1**, plus **cross-embodiment** tests where a **human-view goal image** guides a different robot embodiment after task post-training. Treat those videos as **authors' demonstrations** until independent labs reproduce them.\n\n## How to try it this week\n\n**Read and cite**\n\n- Paper PDF: [arxiv.org/abs/2611.06817](https://arxiv.org/abs/2611.06817?ref=explainx)\n- Overview and films: [dagroup-pku.github.io/open-humanwam/](https://dagroup-pku.github.io/open-humanwam/?ref=explainx)\n\n**Download**\n\n- Models collection: `huggingface.co/collections/DAGroup-PKU/open-humanwam`\n- Dataset: `huggingface.co/datasets/DAGroup-PKU/Open-HumanWAM-50K`\n- Code: `github.com/DAGroup-PKU/HumanNet` branch`open-humanscale`\n\n**Evaluate**\n\n- Start in **RoboTwin 2.0** with their published post-training recipe before attempting real robots.\n- Log **Clean vs Random** separately; Random is where[Xiaomi U0-style synthesis](https://explainx.ai/blog/xiaomi-robotics-u0-world-foundation-model-embodied-synthesis-july-2026) and domain-randomization papers claim the real generalization win.\n\n**Compute expectations**\n\nFifty thousand hours of video is not a laptop fine-tune. Plan cluster storage for the corpus subset you need, mirror HF artifacts regionally, and budget GPU time for world-model pretraining at the scales their curves imply—100-hour baselines are useful sanity checks, not the headline model.\n\n## RoboTwin 2.0 and RoboDojo in plain language\n\n**RoboTwin 2.0** is a two-arm manipulation benchmark with **fifty tasks**, each trained from tens of clean demonstrations, then tested in **matched clean scenes** and **randomized scenes** where object poses, textures, and lighting shift. Papers report **Clean2Clean** and **Clean2Random** separately because randomized success collapses when policies memorize a single tabletop layout.\n\nOpen-HumanWAM's **~80% vanilla / ~89% goal** averages are **post-trained** numbers: they assume you run the authors' adaptation stage on RoboTwin demonstrations after human-video pretraining. The **13.94% zero-shot** figure for HumanWAM-Goal is a different protocol—evaluating the pretrained checkpoint **before** task-specific post-training—so it should not be compared directly to the eighty-eight percent line in the same breath.\n\n**RoboDojo** metrics in the paper (mid-teens percent success in the HF abstract) act as a secondary stress test. Treat them as **additional evidence**, not the headline marketers will tweet.\n\n## World–action models versus classic VLAs\n\nClassic **vision-language-action (VLA)** stacks map an image and instruction directly to joint targets or end-effector deltas. **World–action models (WAMs)** also model short-horizon **dynamics** or **video prediction** jointly with actions, which can help when contact-rich physics matter.\n\nHumanWAM's architecture diagrams on the project site emphasize **closed-loop simulation evaluation** during scaling studies—that is, the model is not judged only on next-token loss on human YouTube clips. For builders comparing stacks, our [LLM vs CLM vs decision models guide](https://explainx.ai/blog/llm-vs-clm-vs-decision-models-which-to-use-2026) is about language heads, not robot WAMs, but the same lesson applies: **match the evaluation harness to the deployment harness** before choosing a pretrained initialization.\n\nIf you already run a **π₀-class diffusion policy** or a smaller **8B Verne checkpoint**, treat HumanWAM as a **pretraining initializer** experiment: swap the human-video encoder weights, keep your robot data collection pipeline fixed for a month, and measure **data efficiency** (demos needed to hit 70% on your task) rather than chasing leaderboard points on tasks you will never ship.\n\n## What people are asking on Hugging Face\n\nThe author comment on the [paper page](https://huggingface.co/papers/2611.06817?ref=explainx) summarizes HumanWAM-Vanilla at **92.08% / 68.88%** Clean/Random (**~80.48%** average) and HumanWAM-Goal at **92.70% / 84.60%** (**~88.65%** average) after post-training, plus **RoboTwin rank #1** language in their self-reported comparison.\n\nA community comment from **PolyHelperAI** notes the **log-linear** hours-to-success relationship and asks where the curve **flattens**—a fair open question the paper's appendices may address with larger held-out suites. Treat flattening points as **empirical**, not guaranteed log-linear forever.\n\nThe automated **Librarian Bot** thread links adjacent 2026 papers (IronMind, InternW0, Ego-CEVR, and others). Readers building literature reviews should cluster those works on **egocentric pretraining** rather than reading Open-HumanWAM in isolation.\n\n## Limits and honest comparisons\n\n**Simulation-first.** Most cited percentages are **RoboTwin 2.0** or simulation analogs. Real-robot zero-shot averages in the teens-to-thirties percent range on the project site are **far from production reliability**.\n\n**Naming drift.** Checkpoints use the **HumanWAM** prefix; the Git branch is `open-humanscale` while Hugging Face collections say **open-humanwam**. The corpus card is **Open-HumanWAM-50K** everywhere official—do not confuse it with the separate HumanNet megacorpus roadmap.\n\n**Leaderboard volatility.** [Verne's model cards](https://explainx.ai/blog/verne-robotics-v0-5-open-weight-world-action-model-8b-libero-99-3-robotwin-81-gated-license-claim-check-2026) already showed rivals ahead on some RoboTwin sorts while marketing claimed leadership. Always read **which split, which post-training budget, and which open-vs-closed row** a table includes.\n\n**Data ethics and consent.** Fifty thousand hours of egocentric human video raises **privacy and consent** questions the technical paper may only partially document. Teams fine-tuning in the EU or California should trace **dataset provenance** fields on Hugging Face before merging clips into commercial robot products.\n\n## Where this fits the embodied foundation model map\n\nOpen-HumanWAM sits in the **open recipe** camp with GreenWAM-style releases ([Sber GreenWAM](https://explainx.ai/blog/greenwam-sber-robotics-world-foundation-model-20-robot-families-2026)) and closed-stack giants shipping APIs only. The differentiator is **reproducible scaling science**: curves tying hours to success, ablations on diversity, and zero-shot goal conditioning you can inspect—not only a mean success scalar.\n\nFor practitioners, the actionable bet is empirical: **if your robot line already pays for human video**, compare fine-tuning **with and without** a HumanWAM initialization before commissioning another year of teleop. If you are simulation-only, rerun their **100-hour vs 1k-hour** checkpoints to see whether log-linear gains appear on **your** end-effector and camera mount.\n\nIndustrial labs chasing [reward-guided OM-1 style human data](https://explainx.ai/blog/reward-ai-om-1-robot-foundation-model-human-data-2026) should watch whether DAGroup publishes **training cost envelopes** in later revisions; without them, total cost of ownership vs proprietary APIs stays fuzzy.\n\n## Related reading\n\n*Benchmark numbers, dataset availability, and repository branches reflect DAGroup-PKU releases as of November 9, 2026.*", "url": "https://wpnews.pro/news/open-humanwam-50k-hours-of-egocentric-video-for-robot-wams", "canonical_source": "https://explainx.ai/blog/open-humanscale-50k-humanwam-robotwin-egocentric-pku-2026?utm_source=rss&utm_medium=feed&utm_campaign=blog", "published_at": "2026-11-09 00:00:00+00:00", "updated_at": "2026-11-09 05:25:11.266163+00:00", "lang": "en", "topics": ["robotics", "ai-research", "machine-learning", "artificial-intelligence"], "entities": ["Peking University", "DAGroup", "Open-HumanWAM", "HumanWAM-Vanilla", "HumanWAM-Goal", "RoboTwin 2.0", "Hugging Face", "RoboDojo"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/open-humanwam-50k-hours-of-egocentric-video-for-robot-wams", "markdown": "https://wpnews.pro/news/open-humanwam-50k-hours-of-egocentric-video-for-robot-wams.md", "text": "https://wpnews.pro/news/open-humanwam-50k-hours-of-egocentric-video-for-robot-wams.txt", "jsonld": "https://wpnews.pro/news/open-humanwam-50k-hours-of-egocentric-video-for-robot-wams.jsonld"}}