orbit

Stable and Efficient Reinforcement Learning for Trillion-Parameter LLMs

Sphere-AI-Lab/orbit

View on GitHub ↗

Stable and Efficient Reinforcement Learning for Trillion-Parameter LLMs

orbit grew -1.3% in the last 30 days, ranks 68th of 97 in Reinforcement Learning / RLHF, last commit 72 days ago.

Star trajectory

82 days tracked
2026-08-31155 stars today2026-11-22
Momentum

-1.3% in 30d, below the Reinforcement Learning / RLHF median of +1.0%.

Maintenance

Last commit 72 days ago, staler than the median of 40d.

How it compares in Reinforcement Learning / RLHF

orbit
155
stars
-1.3% 30d
commit 72 days ago
Reinforcement Learning / RLHF median
287
stars
+1.0% 30d
commit ~40d ago
Leader · OpenRLHF
10.2k
stars
+0.6% 30d
commit 19 days ago

Ranked 68th of 97 by stars.

Ranked by shared tags, then keywords, then stars
  1. Halo is an open-source framework built by White Circle for training large language and multimodal models

    +34 (+4.1%) in 30d · Committed today · Python · Other license

    Shares Pretraining & Training, Training Toolkits, Reinforcement Learning / RLHF

  2. Comprehensive toolkit for Reinforcement Learning from Human Feedback (RLHF) training, featuring instruction…

    +1 (+0.5%) in 30d · Committed 30d ago · Python

    Shares Pretraining & Training, Training Toolkits, Reinforcement Learning / RLHF

  3. An Asynchronous Reinforcement Learning Engine for Omni-Modal Post-Training at Scale

    +27 (+4.1%) in 30d · Committed today · Python · Apache-2.0

    Shares Reinforcement Learning / RLHF, Training Toolkits, Pretraining & Training

  4. AxisRL is an agentic RL post-training framework built on SGLang rollout, Megatron training, and real-world…

    +1 (+0.2%) in 30d · Committed 110d ago · Python · Apache-2.0

    Shares Reinforcement Learning / RLHF, Training Toolkits, Pretraining & Training

  5. From-scratch PyTorch replications of classic and SOTA AI/ML papers: transformers, attention, GANs, diffusion…

    Committed 17d ago · Jupyter Notebook · Apache-2.0

    Shares Pretraining & Training, Training Toolkits, Reinforcement Learning / RLHF

  6. MiroRL is an MCP-first reinforcement learning framework for deep research agent.

    -1 (-0.4%) in 30d · Committed 451d ago · Python · Apache-2.0

    Shares Reinforcement Learning / RLHF, Training Toolkits, Pretraining & Training

  7. A Mac-local specialist-model factory for data, post-training, evaluation, packaging, and evidence-backed…

    +0 (+0.0%) in 30d · Committed yesterday · Swift · MIT

    Shares Training Toolkits, Pretraining & Training, Reinforcement Learning / RLHF

  8. 🚀 Ultra Recipe for Training Long-Horizon Search Agents - matching frontier AI's search capability with a 20B…

    +1 (+0.1%) in 30d · Committed 159d ago · Python · Apache-2.0

    Shares Pretraining & Training, Training Toolkits, Reinforcement Learning / RLHF

  9. Official repository for DR Tulu: Reinforcement Learning with Evolving Rubrics for Deep Research

    +7 (+1.0%) in 30d · Committed 43d ago · Python · Apache-2.0

    Shares Reinforcement Learning / RLHF, Training Toolkits, Pretraining & Training

  10. LLM-authored task-specific Jev harnesses with optional full-trajectory reward reflection and GEPA evolution.

    +27 (+5.1%) in 30d · Committed 37d ago · Python

    Shares Pretraining & Training, Reinforcement Learning / RLHF, Training Toolkits

Language
Python
License
Apache License 2.0
Contributors
3
Forks
10
Watchers
4
Open issues
1
Releases
0
Created
2026-05-21
Stars this week
-1

You've seen where orbit stands today.

Get the moves before everyone else: the 5 Reinforcement Learning / RLHF repos climbing fastest, every Tuesday.

Free weekly digest. Double opt-in, unsubscribe anytime.

Maintain orbit?

Claim this page to add your own call-to-action.

Claim this page