Math RLVR recipe

This recipe runs a small math RLVR task with a GSM8K-style dataset loader and math verification reward function.

areno train \
  --ckpt Qwen/Qwen3-0.6B \
  --dataset-path gsm8k:main \
  --dataset-loader-fn examples/math/dataset_loader.py \
  --reward-fn-path examples/math/math_verify_reward.py \
  --algo gspo \
  --tp-size 1 \
  --world-size 1

Key files:

• examples/math/dataset_loader.py normalizes the dataset.

• examples/math/math_verify_reward.py scores completions.

• Training CLI reference documents rollout, loss, and checkpoint flags.

Adapt this recipe by replacing the dataset loader and reward function first. Keep the batch size small until the environment and reward signal are stable.