Thanks for your great work. I noticed you mentioned the noise in goal specifications and conducted experiments on it.
The Gaussian noise for Image Instance Navigation is relatively easy to implement. But the uncommon synonyms and different descriptions used in the paper are hard to obtain. Thus, I wonder if it's possible to have access to the noisy data you used in GOAT-bench.
Thanks for your great work. I noticed you mentioned the noise in goal specifications and conducted experiments on it.
The Gaussian noise for Image Instance Navigation is relatively easy to implement. But the uncommon synonyms and different descriptions used in the paper are hard to obtain. Thus, I wonder if it's possible to have access to the noisy data you used in GOAT-bench.