In this post I will be discussing about a paper on controlled text generation which was recently accepted at ICLR 2021 main conference track.

In this post I discuss about a paper which studies about the scenarios where loss based prioritization method fails to account when there are noisy / corrupted examples in the training data by inducing noise in the training data.

In this blog, I disuss about the itenary, activities that encompassed our trip to Kedarkantha peak. This marks my second trek to himalayas and the first trek in the state of Uttharakand The land of Gods. Do read if you want to live through the experience of Bungee jumping, rafting and hiking in a single post :).

In this post I will be discussing about a paper on new paradigm which when applied to Transformer based Sequence-to-sequence models has equalled or exceeded the SOTA on 12 downstream summarization datasets. It also has worked well on low resource abstractive summarization tasks.

In this post I will be talking about a paper on abstractive summarization which uses a generator network to generate new words when extracting summaries via a vanilla Seq2seq model with attention. This paper at the time (2017) out performed the SOTA on ROUGE by atleast 2 points.