Researchers release 3,718 labelled robot episodes from ArmnetBench evaluation

Published Tuesday, November 3, 2026 · 2:15 AM EST

Researchers released 3,718 real-robot episodes from the ArmnetBench v0.1 evaluation on the Hugging Face Hub, according to a preprint. The corpus includes 600 teleoperated demonstrations and 3,118 rollouts from seven policies tested on 12 tabletop tasks using low-cost SO-101 robot arms. Each episode contains three camera views, joint states and actions, with data available in LeRobot v3.0 and a reward-model format. An operator labelled each rollout successful, suboptimal or failure; 73% of rollouts are failures. For each task, every policy used the same 50 training demonstrations and ran on the same cell, providing failures across policies under shared conditions.

Sources

arXiv Query: search_query=cat:cs.RO&id_list=&start=0&max_results=30

In this story

Hugging Face

Published by Tech & Business, a media brand covering technology and business. This story was sourced from arXiv Query: search_query=cat:cs.RO&id_list=&start=0&max_results=30 and reviewed by the T&B editorial agent team.

Back to Newswire

Keep reading

Full wire

AI Products

Nvidia reportedly requires cloud clients to direct 51%+ of services outside China

Nvidia verbally told cloud computing clients to sell 51%+ of their services to non-Chinese companies, the Financial Times reported, citing sources.

Science AI

SealCall preprint demonstrates encrypted safety checks for AI agent tool calls

Researchers report a framework that lets external providers audit AI agent tool calls without receiving plaintext audit inputs.

AI Science

Preprint demonstrates correction for unintended effects of language model steering

Researchers report in a preprint that a second steering vector corrects about 78% of unintended changes in generated text while preserving the intended behavior in about 93% of generations.

Products

Ibotta revenue rises 17% as third-party redemptions expand

Ibotta reported third-quarter revenue of $97.8 million, up 17% year over year, as total redeemers reached 23.6 million, according to GuruFocus.

Robotics

Toyota and Waymo unveil bZ4X robotaxi for Japan testing in 2027

Toyota and Waymo unveiled a bZ4X-based robotaxi that Waymo plans to test in Japan beginning in 2027, Yahoo News UK reported. Toyota will install and integrate Waymo's autonomous-driving equipment at its Takaoka factory in Toyota City, Aichi Prefecture.

Security

Attackers target unpatched WordPress payment flaw that exposes wallet credentials

Attackers are attempting to exploit an unpatched vulnerability in the CryptoPayment Gateway WordPress plugin, Defused Cyber said. CVE-2026-81648 has a CVSS score of 10.0 and affects versions 1.2.1 through 1.2.2, the latest version.