AI Paper Atlas

Research map / Reinforcement learning / Imitation learning

Dialogue Generation: From Imitation Learning to Inverse Reinforcement Learning

Ziming Li, Julia Kiseleva, Maarten de Rijke

AAAI 2019 · Reinforcement learning › Imitation learning · approach: Adversarial imitation learning

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Use a state-action reward model with importance sampling instead of the DG-AIL discriminator reward (ablation)DG-AIRL dialogue generation · MovieTriples human pairwise vs DG-AILWin rate: 0.36 win, 0.37 tie, 0.27 lossMixed
Add causal entropy regularization to adversarial training (ablation)SeqGan · MovieTriples human pointwiseAverage score: 0.39 -> 0.53Improved
Add causal entropy regularization to adversarial training (ablation)SeqGan · MovieTriples human pairwise vs DG-AIRLWin rate: 0.28 loss -> 0.27 lossImproved
Replace the adversarial imitation reward with a learned state-action reward and importance sampling (ablation)DG-AIL · MovieTriples testAverage embedding: 0.553 ± 0.003 -> 0.589* ± 0.003Improved
Replace the adversarial imitation reward with a learned state-action reward and importance sampling (ablation)DG-AIL · MovieTriples human pointwiseAverage score: 0.53 -> 0.54Improved
Replace the adversarial imitation reward with a learned state-action reward and importance sampling (ablation)DG-AIL · MovieTriples human pairwiseWin rate: 0.36 win, 0.37 tie, 0.27 lossMixed
Use DG-AIRL instead of Seq2SeqDialogue generation · MovieTriples human pairwiseWin rate: 0.44 win, 0.29 tie, 0.27 lossImproved
Use DG-AIRL instead of VHREDDialogue generation · MovieTriples human pairwiseWin rate: 0.46 win, 0.32 tie, 0.22 lossImproved
Use DG-AIRL instead of SeqGanDialogue generation · MovieTriples human pairwiseWin rate: 0.47 win, 0.25 tie, 0.28 lossImproved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy