data release

 Hi, thanks for the great work! I'm trying to reproduce the Table 1 results.

  Could you share the pre-computed embedding files (train_100k.npy, multi_response_embeddings.npy,
  multi_response_rewards.npy)? If not, could you provide details on how to extract them (e.g., which model,
   which layer, pooling method)?

  Thanks!