AI Paper Atlas

Research map / Large language models / Agents, code and math

Parsel🐍: Algorithmic Reasoning with Language Models by Composing Decompositions

Eric Zelikman, Qian Huang, Gabriel Poesia, Noah Goodman, Nick Haber

NeurIPS 2023 · Large language models › Agents, code and math · approach: LLM-guided program synthesis · 0 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
Decompose tasks into Parsel functions and search combinations of their implementations with testsCode generation on APPS competition-level · APPS competition-levelpass rate: 14.5% to 25.5%Improved
Translate high-level plans directly to Python instead of using Parsel synthesis (ablation)Codex code generation · APPS competition-level, 100 randomly sampled problemspass rate: 25.5% to 6%Hurt
Use Parsel decomposition and synthesis rather than direct GPT-4 generationGPT-4 code generation · HumanEvalpass@1: 67% to 85%Improved
Use Parsel with generated tests and CodeT selection instead of GPT-4 with CodeT aloneGPT-4 code generation · HumanEval, generated testspass rate: 81.1% to 85.1%Improved
Generate combinations of function implementations with Parsel rather than direct programsGPT-4 code generation · HumanEval, pass@anypass rate: 82.3% to 91.5%; 84.1% to 93.9%Improved
Present plans generated with Parsel rather than the baseline planning approachVirtualHome robotic planning · VirtualHome, human accuracy rankingscomparisons preferred over baseline: 69%Improved
Generate plans with Parsel rather than the baseline planning approachVirtualHome robotic planning · VirtualHome, human clarity rankingsrelative likelihood of preference: 70% more likelyImproved
Show indented function names alongside step-by-step output (ablation)Parsel robotic plans · VirtualHome, human accuracy and clarity rankingspreference relative to non-indented ParselNo clear effect
Increase the number of program evaluations for generated function combinationsParsel · APPS competition-levelpass rateImproved
Use Parsel composition to combine function implementations instead of sampling only complete programsCode generation on chained-component tasks · Chained componentsproblems solved: 1x2; 16 programsImproved
Use GPT-4 alone instead of GPT-4 with Parsel (ablation)GPT-4 code generation · HumanEval, pass@anyproblems solved: two problemsMixed

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