Joined November 2025
- Tweets 418
- Following 40
- Followers 21,799
- Likes 537
150 Photos and videos
Pinned Tweet
Mar 24
Axis is officially LIVE on @base. 🔵
Axis is scaling Physical AI for the real world, contributed by everyone.
You can control robots in a virtual world, generate training data at scale, and help build the brain behind tomorrow's robots. All from browser. No hardware needed.
Start building robotics intelligence today: hub.axisrobotics.ai
317
265
1,007
227,731
Jun 11
Excited to build robotics on @base.
Real energy at SuperAI with @baseapac.
We’re just getting started bringing robots, AI, and onchain coordination together.
AI is exciting.
AI onchain is even more exciting.
Great to showcase some of the cutting-edge AI robotics innovations being built on @base to @Hassan_NY, Country Director of @CoinbaseSG, at SuperAI.
Featuring @axisrobotics and @InvLambda 🤖🟦
97
50
165
5,782
Jun 10
It's AXIS.
94
55
183
6,461
Jun 10
The largest gain comes from Layout, the axis where structured scene diversity matters most.
LIBERO-Plus evaluates robustness across seven perturbation axes: Camera, Light, Sensor Noise, Background, Layout, Language, and Robot.
The largest absolute gain from AXIS-100% appears on Layout: 23.2 points over vanilla π0.5.
This directly validates one of the core AXIS data augmentation methods of layout randomization.
AXIS also improves more than visual robustness.
Beyond Layout, AXIS-100% also improves several other LIBERO-Plus axes. These axes measure different kinds of distribution shift.
The result shows that AXIS also improves robustness to viewpoint shifts, degraded visual observations, robot initial-state changes, and task wording variation.
37
26
82
2,829
Jun 10
Together, these results suggest that AXIS improves robustness across perception, scene geometry, control initialization, and language-conditioned task execution.
This is why we describe AXIS as a growable data engine rather than a static dataset: more structured, validated, and semantically preserved data continues to translate into measurable policy improvement.
15
4
26
1,572
Jun 10
We start from the released π0.5 checkpoint, optionally pretrain on AXIS or a matched-volume RoboCasa control, then fine-tune on LIBERO and evaluate on LIBERO-Plus.
With AXIS-100% (full dataset), π0.5 improves from 66.5 to 79.4 overall on LIBERO-Plus, a 12.9 point gain.
The matched-volume RoboCasa control uses the same trajectory count as AXIS-100%, but reaches only 57.5 overall, below the vanilla π0.5 baseline.
This suggests that the gain is not driven by data volume alone, but by the AXIS pipeline: broader task coverage, diverse human demonstrations, data cleaning, and large-scale semantic-preserving randomization.
14
7
34
1,616
Jun 10
AXIS dataset shows a clear scaling trend.
We evaluate three AXIS snapshots under the same architecture, training budget, rollout protocol, and success checkers. The only major variable is the pretraining corpus.
Overall LIBERO-Plus success increases monotonically as AXIS grows, and the gap from AXIS-25% to AXIS-100% is still 7.4 points, suggesting the benefit has not saturated at the current scale.
This is exactly what we want from a growable data engine: adding validated data continues to produce measurable policy improvement.
10
4
20
1,191
Jun 10
In our conference submission, we evaluate AXIS as a growable data engine for robot manipulation through three questions:
1. Does AXIS pretraining improve π0.5 on downstream LIBERO-Plus robustness tasks, beyond a matched-volume baseline?
2. Does the gain scale with AXIS data volume, from 25% to 50% to 100% of data volume?
3. Which perturbation axes benefit the most, and do they match the diversity targeted by our augmentation pipeline?
Here, “AXIS” refers to our growable manipulation dataset snapshot built around a Franka Research 3 robot: 207 tabletop tasks across 7 scene categories, 50k human demonstrations, and 60k task/scene variants produced through cleaning and semantic-preserving augmentation.
Findings below 🧵
Jun 8
Axis Weekly
This week was about making the AXIS loop more scalable end to end: automating data-to-model workflows, testing recovery-driven training, expanding TaskGen coverage, and preparing the dataset and model stack for release.
Key updates:
- Data-to-model automation: We used scripts to speed up and standardize several repetitive but critical workflows.
- Continuous-growth training: We completed multi-data-scale training and success-rate comparisons across several failure tasks.
- Failure task expansion: A new batch of failure tasks has been pushed to test, expanding the evaluation range for ablations across data scale, data quality, and randomization.
- TaskGen: Articulated-object generation is now merged into the automatic generation pipeline.
- Model and release prep: We finished the first round of fine-tuning, evaluation, and benchmarking, completed the dataset’s conference submission, and are now improving experimental results for release.
Details below 🧵
63
25
127
4,789
Jun 9
GAxis (✱,✱)☀️
New week. New challenges. New access.
We’re giving away 10 BitRobot access codes over the next 72 hours.
Winners will get access to SN/04 and start earning rewards from both Axis and BitRobot.
To join:
1. Follow @axisrobotics & @BitRobotNetwork
2. Like repost this post
3. Comment with a screenshot or photo of where you’re training right now
Grinding tasks? Climbing the leaderboard?
Show us your journey.
Jun 4
Announcing our collaboration with @BitRobotNetwork!
Axis is launching SN/04 on BitRobot, the open robotics lab on Solana that coordinates distributed contributors to accelerate Physical AI research.
SN/04 is a teleop-in-sim mission where contributors complete web-based robotics simulation tasks, generate valuable training data, and earn rewards from both ecosystems.
Together, we’re scaling human demonstrations for Physical AI — powered by everyone.
Rules and details below ↓
189
142
309
12,961
Jun 8
Axis Weekly
This week was about making the AXIS loop more scalable end to end: automating data-to-model workflows, testing recovery-driven training, expanding TaskGen coverage, and preparing the dataset and model stack for release.
Key updates:
- Data-to-model automation: We used scripts to speed up and standardize several repetitive but critical workflows.
- Continuous-growth training: We completed multi-data-scale training and success-rate comparisons across several failure tasks.
- Failure task expansion: A new batch of failure tasks has been pushed to test, expanding the evaluation range for ablations across data scale, data quality, and randomization.
- TaskGen: Articulated-object generation is now merged into the automatic generation pipeline.
- Model and release prep: We finished the first round of fine-tuning, evaluation, and benchmarking, completed the dataset’s conference submission, and are now improving experimental results for release.
Details below 🧵
Jun 1
Axis Weekly
This week was about trust and transfer: making community data cleaner, generated tasks broader, and trained policies more robust as they move from simulation to real robots.
Key updates:
- Data quality: We completed a suspicious-user audit script to detect abnormal collection behavior using user statistics and replay/verify failure reasons.
- Webapp and simulation: We improved key gripper and asset interactions, including penetration, heavy-object grasping, and IK flexibility.
- Recover-from-failure: We tested Failure Task 892 and collected 300 failure initial states, with the next round moving to repaired and more randomized tasks.
- TaskGen: Articulated-object generation is now merged into the automatic generation pipeline, covering cabinets, dishwashers, drawers, and existing randomization workflows.
- Model and real-world stack: We completed the first round of fine-tuning, evaluation, and benchmarking, merged the π0.5 evaluation pipeline into the real-world stack, and started bringing a new embodiment into the loop.
A closer look at this week’s progress 🧵
128
57
232
13,762
Jun 8
On the model side, we finished the first round of fine-tuning, evaluation, and benchmarking, and are now adjusting the data recipe for better performance.
The π0.5 evaluation pipeline has been merged into the real-world stack, while web policy inference can now load model checkpoints for online deployment.
We also completed the dataset’s conference submission and are now improving experimental results for the upcoming release.
Next, we will continue batch ablations, generate checkpoints and failure tasks at scale, and land model visualization in the hub.
We are also starting to connect the stack to new real-world embodiments. More on this soon.
6
6
26
1,313
Jun 7
1M trajectories generated on Axis.
A major milestone for our distributed Physical AI data engine.
204
98
351
16,603
Axis Robotics retweeted
Jun 3
For example, the images below show a real on-site deployment scenario at one of our manufacturing customers.
If we want to help them achieve a scalable, end-to-end “pick-anything” solution, the process is non-trivial. We need to collect large-scale data across their diverse SKUs and part geometries, train a model that can generalize across variations, and then deploy it onto the real production line.
From there, we continue collecting real-world feedback data to fine-tune the policy in production. The goal is to progressively improve stability, cycle time, and accuracy — ultimately delivering a system that is faster, more reliable, and production-ready.
Jun 3
I'm very interested in the Task Packages section. Does the team have any demos or detailed documentation for manufacturing businesses? I'd like to learn more.
This closed-loop system solves a major pain point regarding data quality. Wishing the team continued success in compound moat!
40
29
95
5,737
Jun 5
5/6
From a product and strategy perspective, Axis’s data capability is expanding in two directions at the same time.
On one hand, we continue to improve data diversity by covering more scenes, assets, layouts, and visual variations.
On the other hand, we are also increasing task complexity, pushing the data toward longer-horizon, higher-level behaviors that require more reasoning, coordination, and recovery.
14
6
38
1,588
Jun 5
6/6
In short, we are building a more valuable data distribution: from single-step actions to multi-stage tasks, from single-arm manipulation to bimanual coordination, and from a single robot embodiment to cross-embodiment adaptation.
This direction brings us closer to a data infrastructure capable of supporting more complex, realistic, and generalizable robotic intelligence.
9
4
29
1,348