A Better Variant of Self-Critical Sequence Training

Abstract

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.

References in corpus (3)

• A Deep Reinforced Model for Abstractive Summarization
• Variational inference for Monte Carlo objectives
• Context-Aware Visual Policy Network for Sequence-Level Image Captioning

Cited by in corpus (8)

• How Much Can CLIP Benefit Vision-and-Language Tasks?
• Reinforcement Learning-powered Semantic Communication via Semantic Similarity
• Embedded Heterogeneous Attention Transformer for Cross-lingual Image Captioning
• End-to-End Supermask Pruning: Learning to Prune Image Captioning Models
• Exploiting Multiple Sequence Lengths in Fast End to End Training for Image Captioning
• Semi-Autoregressive Transformer for Image Captioning
• Goal-driven text descriptions for images
• ReGen: Reinforcement Learning for Text and Knowledge Base Generation using Pretrained Language Models