Embodied AI & Robotics Papers Monitor (arXiv)
Pricing
from $2.00 / 1,000 paper records
Track new arXiv papers on embodied AI and robot learning: VLA models, manipulation, humanoids, legged locomotion, sim-to-real, world models. Topic tags, code links, only new papers each run.
Pricing
from $2.00 / 1,000 paper records
Rating
0.0
(0)
Developer
Simon Zhong
Maintained by CommunityActor stats
0
Bookmarked
2
Total users
1
Monthly active users
2 days ago
Last modified
Categories
Security
Limited permissionsShare
Every new paper in these arXiv categories is included (no keyword needed). Default cs.RO = Robotics. Leave empty to rely only on the keyword-gated categories below.
[ "cs.RO"]Papers from these categories are included only if their title or abstract matches one of the embodied-AI keywords below. Default: cs.AI, cs.CV, cs.LG (where many VLA, world-model and robot-learning papers are filed). Other useful ones: cs.CL, cs.MA, eess.SY.
[ "cs.AI", "cs.CV", "cs.LG"]Phrases matched in title/abstract (any of them). Applied only to the extra categories above, not to the core categories. If empty, extra categories behave like core categories (all papers).
[ "embodied", "robot", "robotic", "robotics", "vision-language-action", "VLA", "humanoid", "legged locomotion", "quadruped", "manipulation policy", "robotic manipulation", "dexterous", "sim-to-real", "imitation learning", "visuomotor", "diffusion policy", "teleoperation"]Keep only papers whose primary (main) arXiv category is one of the core or extra categories. Recommended: drops papers from unrelated fields that are only cross-listed (e.g. a physics paper cross-listed to cs.RO). Filtered papers are not output, not charged and not remembered as seen.
Every paper gets rule-based topic tags. Only papers tagged with at least one selected topic are output. The default selection is the embodied-AI topics (VLA, manipulation, humanoids, legged locomotion, dexterous hands, navigation, imitation/robot learning, world models, sim-to-real, tactile…); classical robotics topics such as drones, autonomous driving, SLAM and planning & control are not selected. Clear the list to get every robotics paper. Filtered papers are free.
[ "VLA", "Embodied agents", "Manipulation", "Dexterous hands", "Humanoid", "Legged locomotion", "Navigation", "Imitation learning", "Robot learning", "Diffusion / flow policies", "World models", "Sim-to-real", "Tactile sensing"]Output only papers whose abstract or comments link to GitHub, GitLab, Hugging Face or similar. Filtered papers are free.
Optional phrases that every paper must match (title/abstract by default), e.g. "humanoid", "diffusion policy", "tactile". Combined with Keyword logic below. Leave empty to get everything.
[]How the optional extra keywords above are combined. They are always AND-ed with the category clause.
Where keywords are matched: title + abstract (recommended), title only, or all fields (also authors, comments, journal ref).
Only papers in the last N days (UTC). Ignored if Date from / Date to is set. 0 = no date filter.
Optional explicit window start (UTC, inclusive). Overrides Last N days.
Optional explicit window end (UTC, inclusive). Defaults to now.
Which date the window applies to: first submission (recommended for new-paper monitoring) or last update.
Maximum number of papers to output in this run. Each output paper is one paper-record event ($0.002, i.e. $2 per 1,000). Papers skipped as already seen are free.
Sort order of arXiv results. Use submittedDate for monitoring; relevance is useful for one-off keyword searches.
Descending returns the newest first.
Monitoring mode: skip papers already stored in the 'seen IDs' key-value store. Skipped papers are not charged. Use with a schedule.
With 'Only new', re-output a paper when a higher version appears.
Named key-value store that keeps seen arXiv IDs across runs. Use different names for independent monitors. Empty = do not persist.
When 'Only new', 'Primary category only' or a topic/code filter is on, upper bound of arXiv results scanned to find enough matching papers (each page of 100 takes ≥3 s due to arXiv rate limits).