Axis Robotics

Product Update 🛠️

We've reduced the number of open task slots. Here's why.

1️⃣ GPT-6 can now solve simple tasks and generate reliable trajectories for them. We'll further reduce repetitive collection work as we move forward.
2️⃣ High-quality human demonstrations remain essential https://t.co/nCXR4P1m4J

Product Update 🛠️

We've reduced the number of open task slots. Here's why.

1️⃣ GPT-6 can now solve simple tasks and generate reliable trajectories for them.
2️⃣ High-quality human demonstrations remain essential for complex tasks. These demonstrations carry far more value than https://t.co/UuTVBuIGQz

7/7 Links

Explore AXIS-Bench and our broader data resources:

Project Resources:
Project page: https://t.co/9pvWoFdRzu
Paper: https://t.co/6Irf55g2d5
Code: https://t.co/ONkJ4k41qL
Try Axis Hub: https://t.co/brRVGQrYVQ

Datasets:
Axis Dataset V2: https://t.co/se54ZroQe4 https://t.co/BdFhBbhaxj

6/7 Data Release

Axis Dataset V1 launched in August. We've since built Dataset V2, now covering 1,092 tasks and 2.35M episodes (including camera augmentation). The companion datasets used in AXIS-Bench are all curated from the V2 pool:

• Pretraining dataset. 5,000 simulated https://t.co/U7FM0ja7T3

5/7 Sim to Real

AXIS-Bench also includes a reference workflow and baseline for real-world experiment: align the simulated cameras and workspace, re-render recorded demonstrations, and fine-tune on a small set of real demonstrations.

On a Franka FR3, plain BC pretraining on https://t.co/RJGbCqqnNA

4/7 Key Findings

• How quality is used matters. Approach-aware episode conditioning improves mean success over plain behavioral cloning (BC) across all three VLAs: +3.9%, +23.4%, and +5.6% mean relative gains over three training seeds (π₀.₅, SmolVLA, X-VLA).

• Granularity https://t.co/PWoi9hpmnt

3/7 Experiment Setup

We compare three quality-aware approaches: weighting, filtering, and conditioning, across two reward definitions (object-centric and approach-aware) and two label granularities (episode and action chunk).

The controlled study uses 5,000 simulation https://t.co/ATyOleyayw

2/7 The Idea

Human demonstrations often contain hesitation, detours, and uneven progress. But should a VLA discard suboptimal segments, downweight them, or learn to condition on their quality?

AXIS-Bench builds on crowdsourced simulation demonstrations with recorded object and https://t.co/LjrXBsIkl8

How should VLAs learn from naturally suboptimal robot demonstrations?

Introducing AXIS-Bench: What Matters in Quality-Aware VLA Pretraining, a benchmark and empirical study of how demonstration quality should be defined, labeled, and used during pretraining.

Across three VLAs https://t.co/0h0qloxBEL