A Better Variant of Self-Critical Sequence Training

22 Mar 2020  ·  Ruotian Luo ·

In this work, we present a simple yet better variant of Self-Critical Sequence Training. We make a simple change in the choice of baseline function in REINFORCE algorithm. The new baseline can bring better performance with no extra cost, compared to the greedy decoding baseline.

PDF Abstract

Datasets


Results from the Paper


Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Image Captioning COCO Captions Transformer_NSC BLEU-4 39.4 # 24
METEOR 28.9 # 20
ROUGE-L 58.7 # 8
CIDER 129.6 # 25
SPICE 22.8 # 21
BLEU-1 80.7 # 7
BLEU-2 65.6 # 2
BLEU-3 51.3 # 2

Methods