AI Paper Atlas

Research map / Large language models / Language, safety and interpretability

Joint Prompt Optimization of Stacked LLMs using Variational Inference

Alessandro Sordoni, Xingdi Yuan, Marc-Alexandre Côté, Matheus Pereira, Adam Trischler, Ziang Xiao, Arian Hosseini, Friederike Niedtner, Nicolas Le Roux

NeurIPS 2023 · Large language models › Language, safety and interpretability · approach: LLM pruning and scaling data · 5 citations (OpenAlex)

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
optimize a single-layer prompt rather than use the strongest DaVinci-003 baselineDLN-1 · Hyper. testaccuracy: 68.5{ $\pm$5.5} -> 91.9{ $\pm$3.0}Improved
optimize a single-layer prompt rather than use the strongest DaVinci-003 baselineDLN-1 · Trec testaccuracy: 80.0 -> 89.7{ $\pm$3.2}Improved
optimize a single-layer prompt rather than use the strongest DaVinci-003 baselineDLN-1 · Subj testaccuracy: 76.4 -> 83.2{ $\pm$6.5}Improved
optimize a single-layer prompt rather than use the strongest DaVinci-003 baselineDLN-1 · Nav. testaccuracy: 67.3{ $\pm$7.7} -> 68.5{ $\pm$4.7}Improved
optimize a single-layer prompt rather than use the strongest DaVinci-003 baselineDLN-1 · Logic.7 testaccuracy: 45.5{ $\pm$4.7} -> 47.5{ $\pm$2.1}Improved
optimize a single-layer prompt rather than use the zero-shot baselineDLN-1 · Date. testaccuracy: 56.4 -> 55.7{ $\pm$4.5}Hurt
add a jointly optimized second LLM layerDLN-1 · Nav. testaccuracy: 68.5 -> 83.1Improved
add a jointly optimized second LLM layerDLN-1 · Date. testaccuracy: 55.7 -> 75.2Improved
add a jointly optimized second LLM layerDLN-1 · Logic.7 testaccuracy: 47.5 -> 45.7Hurt
add a jointly optimized second LLM layerDLN-1 · Disaster testaccuracy: 81.7 -> 82.8Improved
add a jointly optimized second LLM layerDLN-1 · Subj testaccuracy: 83.2 -> 85.9Improved
train layerwise and keep the second layer fixed instead of end-to-end training (ablation)DLN-2 · Nav. testaccuracy: 73.1 vs 83.1Hurt

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