Skip to content
youwyuPublic

About

CoRL 2026 🤠

Resources

Stars

1 star

Watchers

0 watching

Forks

Latest commit

 

History

3 Commits

Folders and files

NameName
Last commit message
Last commit date
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

FOSA

Generative Flow on Symbolic Abstractions for Embodiment Capability Discovery

CoRL 2026 Project Website arXiv Hugging Face dataset

Robots training on tasks discovered by FOSA

FOSA discovers which tasks a robot can learn. A GFlowNet over symbolic task skeletons proposes a task, an LLM compiles it into a MuJoCo Playground environment, and PPO trains a task policy on it: the policy's learning curve is the GFlowNet's reward. The task policies are then distilled into one language-conditioned student.

🤖 Installation

conda create -n fosa python=3.11 -y
conda activate fosa
conda install conda-forge::uv -y
uv pip install -e . --no-cache

The robot models are downloaded from the dataset on first use.

LLM

FOSA talks to any OpenAI-compatible chat endpoint. The paper uses Gemma 4 31B, which you can serve locally, e.g. with vLLM:

vllm serve google/gemma-4-31B-it --served-model-name gemma-4-31B-it --port 8000

or use a hosted model: awesome-free-llm-apis lists free ones.

export FOSA_LLM_ENDPOINT=http://localhost:8000/v1/chat/completions  # default
export FOSA_LLM_MODEL=gemma-4-31B-it                                  # default
export FOSA_LLM_API_KEY=...                                           # if the server needs one
Task grammars of the paper

Each robot's task grammar is written once by the LLM. To reuse the paper's, copy them into the log directory:

mkdir -p logs/gemma-4-31B-it && cp -r data/grammars/* logs/gemma-4-31B-it/

TODO

  • Bump mujoco version to 4
  • Full VLA ckpts for all robots

🔥 Discover tasks

Important

The paper runs each robot and method on 4 GPUs (NVIDIA A100, 40 GB).

Robots: go2 (quadruped), jackal (vehicle), booster (humanoid), inspire_hand (hand).

One method at a time

FOSA

python -m fosa.discovery.fosa --robot go2 --gpu-ids 0,1,2,3

Ablations

python -m fosa.discovery.fosa --robot go2 --gpu-ids 0,1,2,3 --method fosa_rnd  # FOSA-RND
python -m fosa.discovery.fosa --robot go2 --gpu-ids 0,1,2,3 --method fosa_llm  # FOSA-LLM

Baselines

python -m fosa.discovery.mcts --robot go2 --gpu-ids 0,1,2,3 --method mcts  # also llm_mcts, muzero (MCTS-Learn)
python -m fosa.discovery.curriculum --robot go2 --gpu-ids 0,1,2,3          # LLM-CL

Each run trains 256 tasks; run the same command again to resume an interrupted run.

Everything at once on Slurm

To run every robot and method of the paper automatically, one job each:

scripts/slurm.sh go2,jackal,booster,inspire_hand all 256

🎓 Distill a student

python -m fosa.learning.dagger --robot go2 --method fosa --gpu-ids 0,1,2,3

On Slurm (the paper uses seeds 1, 2 and 3):

FOSA_STUDENT_SEED=1 scripts/slurm_student.sh go2,jackal,booster,inspire_hand all

📊 Evaluate

Held-out tasks

FOSA's students from the paper are in checkpoints/<robot>/student.pt.

python -m fosa.evaluation.heldout prepare go2   # compile the held-out tasks
python -m fosa.evaluation.heldout train go2     # train a reference policy per task
python -m fosa.evaluation.heldout evaluate go2 --checkpoint checkpoints/go2/student.pt

Citation

@inproceedings{yu2026generative,
  title={Generative Flow on Symbolic Abstractions for Embodiment Capability Discovery},
  author={Youwei Yu and Lantao Liu},
  booktitle={10th Annual Conference on Robot Learning},
  year={2026},
  url={https://openreview.net/forum?id=weyfPS89NJ}
}

About

CoRL 2026 🤠

Resources

Stars

1 star

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages