Everything can be an RL environment now
Building good RL environments is hard.
We're introducing a way to turn anything into RL-environments with REA using the open source AgentEnv framework from @ScaleAILabs
Any site, app, or game can now easily be turned into an env 😉
2
32
4,107
Learn more at agentenvframework.com
1
80
Teaser for the upcoming AirOps RL Environment. A high fidelity environment testing whether a model can run an airline during a snowstorm.
1
16
11,630
Preview available on huggingface.co/datasets/eara…. In collaboration with @adithya_s_k & team. Built using agentenvframework.com
130
Edgar Arakelyan retweeted
that virtual clock fix is so underrated. nothing worse than rl runs breaking because time drifted.
1
49
Try it:
pip install agentenv-framework agentenv-hf
agent-env hf run earakely-scale/PortSimEnv-AgentEnv
It shows what the environment needs and asks before it runs anything.
Open source from @ScaleAILabs: agentenvframework.com
1
2
121
AgentEnv on the Hub, beside Harbor, Verifiers, OpenEnv and NeMo Gym: huggingface.co/datasets?libr…
PortSimEnv: huggingface.co/datasets/eara…
AirOpsEnv: huggingface.co/datasets/eara…
OpenCiv3: huggingface.co/datasets/eara…
PortSimEnv builds on @adithya_s_k's original in FineEnvs.
92
Edgar Arakelyan retweeted
we got lamport clocks for agents before gta 6
Three agents and eight servers ask what time it is, all at once, and every one gets the same answer.
At @ScaleAILabs, we built this virtual clock into our open-source AgentEnv Framework so RL environments stay reproducible. Here's why it matters and how it works 🧵
1
3
214
Three agents and eight servers ask what time it is, all at once, and every one gets the same answer.
At @ScaleAILabs, we built this virtual clock into our open-source AgentEnv Framework so RL environments stay reproducible. Here's why it matters and how it works 🧵
5
5
32
11,666
4. One task step
Putting a task on the virtual clock is one sync_env_clock step, right before the prompt. It arms the clock and syncs the servers that follow it, so virtual time starts with the agent.
1
1
73
Try it: pip install agentenv-framework
A server follows the clock in a few lines. Full post: agentenvframework.com/blog/r…
Code: github.com/scaleapi/agentenv…
Written with @RazvanDuu, @thetututrain, Mateusz Cikowski and @liuying04. How do you handle time in your RL envs?
3
123
Edgar Arakelyan retweeted
The future of enterprise AI is multi-model, and enterprises deserve the freedom to choose.
NEWS: We’re expanding that choice through our collaboration with @NVIDIA, post-training its open-weight Nemotron 3.5 Super VL for enterprise workflows. Here’s what it delivered for two of our customers:
✅ @Skydance: A complex finance workflow that reviews vendor credit memos, statements and invoices to find credits the company is owed. Post-trained Nemotron matched their current frontier model’s accuracy at 40x lower inference cost, with 5x faster response times.
✅ @EY_US: An M&A due diligence agent that finds and reconciles answers across hundreds of files in a virtual data room. Post-trained Nemotron matched EY’s current frontier model at about one-eighteenth of the inference cost, returning answers more than twice as fast.
Put NVIDIA’s open-weight models to work for your business with Scale Post-Training.
Thank you, @JensenHuang and the NVIDIA team, for the early access and collaboration.
scale.com/partnership/nvidia
4
7
44
3,335
It was great collaborating with you and your team. @tensorlake sandboxes are a natural fit for RL Environment builders!
We’re partnering with @scale_AI to integrate @tensorlake sandboxes with AgentEnv.
Here’s a real-world environment: the Port of Barcelona, built with AgentEnv and running in @tensorlake sandboxes. A GLM 5.3-based agent achieves a perfect 1.0 score on a task that manages ship movements in the port based on weather conditions.
Docs below 🧵
1
18
3,140
Edgar Arakelyan retweeted
Tensorlake is now a built-in sandbox provider in @scale_AI 's AgentEnv Framework. An open-source framework for building RL environments.
Check out the docs and the demo to see how the Tensorlake sandbox hosts a full environment for a PortSim task.
agentenvframework.com/docs/s…
Added @tensorlake as a sandbox provider to @scale_AI AgentEnv Framework. I tested it with a PortSim task:
An agent runs a port through a real windy week inside two Tensorlake sandboxes. Got a perfect score.
It was super fun to build and a great environment to play with.
5
7
513
Edgar Arakelyan retweeted
By far the coolest RL environment I've seen, what a time to be alive and building
We collaborated with @scaleAI to build a high-fidelity port simulation RL environment where AI agents tackle real-world logistics challenges
4
318
28,769
Build the world agents learn in with @scale_AI AgentEnv and E2B.
Package your environment as an E2B template, and every task run boots a fresh microVM with the same world inside.
Docs: agentenvframework.com/docs/s…
Every RL environment we build runs on AgentEnv.
Today we’re open-sourcing it, so more people can build the worlds agents learn in.
2
4
28
23,063