Air-Decoding: Attribute Distribution Reconstruction for Decoding-Time Controllable Text Generation

• URL: http://arxiv.org/abs/2310.14892v3
• Date: Thu, 2 Nov 2023 03:14:47 GMT
• Title: Air-Decoding: Attribute Distribution Reconstruction for Decoding-Time Controllable Text Generation
• Authors: Tianqi Zhong, Quan Wang, Jingxuan Han, Yongdong Zhang, Zhendong Mao
• Abstract summary: Controllable text generation (CTG) aims to generate text with desired attributes. We propose a novel lightweight decoding framework named Air-Decoding. Our method achieves a new state-of-the-art control performance.
• Score: 58.911255139171075
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Controllable text generation (CTG) aims to generate text with desired attributes, and decoding-time-based methods have shown promising performance on this task. However, in this paper, we identify the phenomenon of Attribute Collapse for the first time. It causes the fluency of generated text to rapidly decrease when the control strength exceeds a critical value, rendering the text completely unusable. This limitation hinders the effectiveness of decoding methods in achieving high levels of controllability. To address this problem, we propose a novel lightweight decoding framework named Air-Decoding. Its main idea is reconstructing the attribute distributions to balance the weights between attribute words and non-attribute words to generate more fluent text. Specifically, we train prefixes by prefix-tuning to obtain attribute distributions. Then we design a novel attribute distribution reconstruction method to balance the obtained distributions and use the reconstructed distributions to guide language models for generation, effectively avoiding the issue of Attribute Collapse. Experiments on multiple CTG tasks prove that our method achieves a new state-of-the-art control performance.

Related papers

• Continuous, Subject-Specific Attribute Control in T2I Models by Identifying Semantic Directions [21.371773126590874]
  We show that there exist directions in the commonly used token-level CLIP text embeddings that enable fine-grained subject-specific control of high-level attributes in text-to-image models. We introduce one efficient optimization-free and one robust optimization-based method to identify these directions for specific attributes from contrastive text prompts.
  arXiv  Detail & Related papers  (2024-03-25T18:00:42Z)
• Text Attribute Control via Closed-Loop Disentanglement [72.2786244367634]
  We propose a novel approach to achieve a robust control of attributes while enhancing content preservation. In this paper, we use a semi-supervised contrastive learning method to encourage the disentanglement of attributes in latent spaces. We conducted experiments on three text datasets, including the Yelp Service review dataset, the Amazon Product review dataset, and the GoEmotions dataset.
  arXiv  Detail & Related papers  (2023-12-01T01:26:38Z)
• Successor Features for Efficient Multisubject Controlled Text Generation [48.37713738712319]
  We introduce SF-GEN, which is grounded in two primary concepts: successor features (SFs) and language model rectification. SF-GEN seamlessly integrates the two to enable dynamic steering of text generation with no need to alter the LLM's parameters. To the best of our knowledge, our research represents the first application of successor features in text generation.
  arXiv  Detail & Related papers  (2023-11-03T00:17:08Z)
• Look-back Decoding for Open-Ended Text Generation [62.53302138266465]
  We propose Look-back, an improved decoding algorithm that tracks the distribution distance between current and historical decoding steps. Look-back can automatically predict potential repetitive phrase and topic drift, and remove tokens that may cause the failure modes. We perform decoding experiments on document continuation and story generation, and demonstrate that Look-back is able to generate more fluent and coherent text.
  arXiv  Detail & Related papers  (2023-05-22T20:42:37Z)
• DisCup: Discriminator Cooperative Unlikelihood Prompt-tuning for Controllable Text Generation [6.844825905212349]
  We propose a new CTG approach, namely DisCup, which incorporates the attribute knowledge of discriminator to optimize the control-prompts. DisCup can achieve a new state-of-the-art control performance while maintaining an efficient and high-quality text generation, only relying on around 10 virtual tokens.
  arXiv  Detail & Related papers  (2022-10-18T02:59:06Z)
• FAST: Improving Controllability for Text Generation with Feedback Aware Self-Training [25.75982440355576]
  Controllable text generation systems often leverage control codes to direct various properties of the output like style and length. Inspired by recent work on causal inference for NLP, this paper reveals a previously overlooked flaw in these control code-based conditional text generation algorithms. We propose two simple techniques to reduce these correlations in training sets.
  arXiv  Detail & Related papers  (2022-10-06T19:00:51Z)
• Classifiers are Better Experts for Controllable Text Generation [63.17266060165098]
  We show that the proposed method significantly outperforms recent PPLM, GeDi, and DExperts on PPL and sentiment accuracy based on the external classifier of generated texts. The same time, it is also easier to implement and tune, and has significantly fewer restrictions and requirements.
  arXiv  Detail & Related papers  (2022-05-15T12:58:35Z)
• Text Revision by On-the-Fly Representation Optimization [76.11035270753757]
  Current state-of-the-art methods formulate these tasks as sequence-to-sequence learning problems. We present an iterative in-place editing approach for text revision, which requires no parallel data. It achieves competitive and even better performance than state-of-the-art supervised methods on text simplification.
  arXiv  Detail & Related papers  (2022-04-15T07:38:08Z)
• Attribute Alignment: Controlling Text Generation from Pre-trained Language Models [46.19190007510232]
  We propose a simple and flexible method for controlling text generation by aligning disentangled attribute representations. In contrast to recent efforts on training a discriminator to perturb the token level distribution for an attribute, we use the same data to learn an alignment function to guide the pre-trained, non-controlled language model to generate texts with the target attribute without changing the original language model parameters.
  arXiv  Detail & Related papers  (2021-03-20T01:51:32Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.