AI Paper Atlas

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

Grammar Prompting for Domain-Specific Language Generation with Large Language Models

Bailin Wang, Zi Wang, Xuezhi Wang, Yuan Cao, Rif A. Saurous, Yoon Kim

NeurIPS 2023 · Large language models › Language, safety and interpretability · approach: LLM unlearning and benchmarks · 23 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
add predicted specialized grammar before the programstandard prompting · GeoQuery few-shotexecution accuracy (%): 81.5 -> 87.5Improved
add predicted specialized grammar before the programstandard prompting · Overnight-Blk few-shotprogram accuracy (%): 29.3 -> 34.8Improved
use grammar prompting instead of linearized derivation tree promptingderivation tree prompting · GeoQuery few-shotexecution accuracy (%): 77.5 -> 87.5Improved
add grammar constraint on predicted grammar (ablation)grammar prompting · Overnight-Blk few-shotexecution accuracy (%): 57.4 -> 60.4Improved
add program constraint on top of grammar constraint (ablation)grammar prompting · GeoQuery few-shotprogram accuracy (%): 67.9 -> 69.6Improved
add constrained decoding to standard prompting (ablation)standard prompting · Overnight-Blk few-shotexecution accuracy (%): 54.7 -> 54.7No clear effect
replace predicted grammar with oracle grammar (ablation)grammar prompting · SMCalFlow few-shotprogram accuracy (%): 50.8 -> 80.0Improved
grammar prompting in retrieval-based ICLstandard prompting · GeoQuery retrievalexecution accuracy (%): 96.8 -> 97.9Improved
grammar prompting on unseen functionsstandard prompting · GeoQuery NewFuncexecution accuracy (%): 63.3 -> 90.8Improved
grammar prompting with different base LLMsstandard prompting · GeoQuery 100-example subsetexecution accuracy (%): GPT-4 85 -> 98; PaLM 2-L 90 -> 87Mixed
grammar prompting for molecule generationstandard prompting · Chain Extendersvalidity (%): 60.3 -> 96.3Improved
grammar prompting for molecule generationstandard prompting · Isocyanatesdiversity: 0.82 -> 0.79Mixed

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