Update README.md

conda activate vagen

# verl

git clone git@github.com:JamesKrW/verl.git

git clone https://github.com/JamesKrW/verl.git

cd verl

pip install -e .

cd ../

# vagen

git clone git@github.com:RAGEN-AI/vagen.git

git clone https://github.com/RAGEN-AI/VAGEN.git

cd vagen

bash scripts/install.sh

```

git checkout release

cd ../vagen

wandb login # login into wandb

# Then, you can run

bash vagen/vagen/examples/release_experiments/gae.sh # rico-gae

bash vagen/vagen/examples/release_experiments/grpo_mask_loss.sh # rico-grpo + loss mask

bash vagen/vagen/examples/release_experiments/grpo.sh # rico-grpo

bash vagen/vagen/examples/release_experiments/mask_gae_mask_loss_bi_level.sh # trico - turn reward

bash vagen/vagen/examples/release_experiments/mask_gae_mask_loss_turnwise_gae.sh # trico - turn reward - bi-level gae + turn-level gae

bash vagen/vagen/examples/release_experiments/mask_gae_mask_loss_turnwise_reward_bi_level.sh # trico

bash vagen/vagen/examples/release_experiments/mask_gae_mask_loss.sh # aico

bash vagen/vagen/examples/release_experiments/mask_gae.sh # aico - loss mask

bash vagen/vagen/examples/release_experiments/mask_loss.sh # aico - gae mask

bash vagen/examples/release_experiments/gae.sh # rico-gae

bash vagen/examples/release_experiments/grpo_mask_loss.sh # rico-grpo + loss mask

bash vagen/examples/release_experiments/grpo.sh # rico-grpo

bash vagen/examples/release_experiments/mask_gae_mask_loss_bi_level.sh # trico - turn reward

bash vagen/examples/release_experiments/mask_gae_mask_loss_turnwise_gae.sh # trico - turn reward - bi-level gae + turn-level gae

bash vagen/examples/release_experiments/mask_gae_mask_loss_turnwise_reward_bi_level.sh # trico

bash vagen/examples/release_experiments/mask_gae_mask_loss.sh # aico

bash vagen/examples/release_experiments/mask_gae.sh # aico - loss mask

bash vagen/examples/release_experiments/mask_loss.sh # aico - gae mask

```

Each run takes ~4 hours to reach 150 steps on 4 H100s. You can decrease testing frequency to speed up training. Training might be unstable due to loss spikes; we recommend restoring from the latest checkpoint when encountering such cases. We will resolve this issue in future work (see roadmap).