On-the-Fly Controlled Text Generation with Experts and Anti-Experts

By Alisa Liu, Ximing Lu, Swabha Swayamdipta, Chandra Bhagavatula, Noah A. Smith, Yejin Choi. 2021

16 citations recorded by the source corpus.

Subfields: Natural Language Processing, Prompting & Reasoning

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Natural Language Processing

1. Nanyun Peng
2. Y-Lan Boureau
3. Kai-Wei Chang
4. Dirk Hovy
5. Yulia Tsvetkov
6. Emily Dinan
7. Emily Sheng
8. Minlie Huang

In Prompting & Reasoning

1. Chujie Zheng
2. Yulia Tsvetkov
3. Minlie Huang
4. Kai-Wei Chang
5. Emily Dinan
6. Nanyun Peng
7. Fei Mi
8. Sachin Kumar

Citations in the corpus

• Societal Biases in Language Generation: Progress and Challenges (2021, methodology): “, 2020), DExperts (Liu et al., 2021)), there is potential for extending these methods to bias mitigation.”
• Anticipating Safety Issues in E2E Conversational AI: Framework and Tooling (2021, background): “Liu et al. (2021) detoxify a language model’s output by upweighting the probabilities of generating words considered unlikely by a second “anti-expert” model that models toxic language.”
• Guiding the Release of Safer E2E Conversational AI through Value Sensitive Design (2022, background): “In addition, work on building safer LLMs explores fine-tuning on curated data (Solaimon and Dennison, 2021) or directly controlling the generations of the model (Dathathri et al., 2019; Liu et al., 2021; Schick et al., 2021; Xu et al., 2020).”
• Process for Adapting Language Models to Society (PALMS) with Values-Targeted Datasets (2021, background): “Previously proposed debiasing methods include [6]’s foundational work to debias word embeddings; [29]’s use of product of experts to train a model to avoid dataset biases; [39]’s human-and-model-in-the-loop technique to better train and evaluate models without toxic and unwanted behavior; [23]’s use”
• COLD: A Benchmark for Chinese Offensive Language Detection (2022, methodology): “The publicly available detector, PerspectiveAPI2, is widely used for toxicity evaluation and contributes to creating safer environments for online communication (Han and Tsvetkov, 2020; Liu et al., 2021).”
• Controlled Text Generation as Continuous Optimization with Multiple Constraints (2021, methodology): “Recent works address this problem with left-to-right autoregressive decoding, and modify the vocabulary distribution at every step directly using attribute classifier probabilities [69, 37], or indirectly via backpropagating gradients through model activations [9].”