Plug and Play Language Models: A Simple Approach to Controlled Text Generation

Authors: Sumanth Dathathri, Andrea Madotto, Janice Lan, Jane Hung, Eric Frank, Piero Molino, Jason Yosinski, Rosanne Liu

ICLR 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Model samples demonstrate control over a range of topics and sentiment styles, and extensive automated and human annotated evaluations show attribute alignment and fluency.
Researcher Affiliation Collaboration Sumanth Dathathri CMS, Caltech Andrea Madotto HKUST Janice Lan Uber AI Jane Hung Uber AI Eric Frank Uber AI Piero Molino Uber AI Jason Yosinski Uber AI Rosanne Liu
Pseudocode No The paper describes the method using mathematical equations and descriptive text, but does not include any clearly labeled 'Pseudocode' or 'Algorithm' blocks.
Open Source Code Yes Code for the experiments is available at: https://github.com/uber-research/PPLM.
Open Datasets Yes The sentiment discriminator here distinguishes sentiment between POSITIVE and NEGATIVE and is trained on the SST-5 dataset (Socher et al., 2013).
Dataset Splits Yes The sentiment discriminator here distinguishes sentiment between POSITIVE and NEGATIVE and is trained on the SST-5 dataset (Socher et al., 2013).
Hardware Specification No The paper does not provide specific details regarding the hardware used for running experiments, such as GPU or CPU models.
Software Dependencies No The paper mentions using a GPT-2 model and a PyTorch transformer, but does not provide specific version numbers for these or other software dependencies.
Experiment Setup Yes Table S18: The full set of hyperparameters used in each task in the experiments section. Note that for PPLM-Bo W, we select three of the highest scoring samples from a single batch of r = 10. For PPLM-Discrim, we get 1 sample per batch, across 3 batches of r = 10. Method Type Attribute Hyperparameters PPLM-Bo W Politics, Legal, Computers, Space, Science, Military m = 3, λkl = 0.01, α = 0.01, γ = 1.5, γgm = 0.9, r = 10, τ = 0.85 PPLM-Bo W Religion m = 3, λkl = 0.01, α = 0.01, γ = 1.5, γgm = 0.8, r = 10, τ = 0.85 PPLM-Discrim POSITIVE, NEGATIVE m = 10, λkl = 0.01, α = 0.03, γ = 1.0, γgm = 0.95, r = 10, τ = 0.9 PPLM-Discrim Detoxicification m = 10, λkl = 0.01, α = 0.02, γ = 1.0, γgm = 0.9, r = 1, τ = 0