AI Paper Atlas

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

ANPL: Towards Natural Programming with Interactive Decomposition

Di Huang, Ziyuan Nan, Xing Hu, Pengwei Jin, Shaohui Peng, Yuanbo Wen, Rui Zhang, Zidong Du, Qi Guo, Yewen Pu, Yunji Chen

NeurIPS 2023 · Large language models › Agents, code and math · approach: LLM-guided program synthesis · 4 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
use ANPL interactive decomposition instead of ChatGPT with interactionGPT-3.5-turbo · ARC 400 taskssolving rate (%): 58.4 -> 75.0Improved
remove interaction from ANPL (one-shot) vs vanilla GPT one-shot (ablation)ANPL · ARC 400 taskssolving rate (%): 16.8 -> 23.5Improved
allow interaction in ANPL vs one-shot ANPL (ablation)ANPL · ARC 400 taskssolving rate (%): 23.5 -> 75.0Improved
allow interaction with ChatGPT vs one-shot GPT (ablation)GPT-3.5-turbo · ARC 400 taskssolving rate (%): 16.8 -> 58.4Improved
ANPL vs ChatGPT when users cannot see Python code (beginners)GPT-3.5-turbo · ARC beginner setting, 40 problemssolving rate (%): 37.2 -> 89.9Improved
ANPL vs ChatGPT in lines of code writtenANPL · ARC 400 tasksmean lines of code: 26.47 -> 10.29Improved
ANPL vs ChatGPT in time waiting for GPTANPL · ARC 400 taskswaiting time (%): 23.5 -> 14.6Improved
ANPL vs ChatGPT in variation across usersANPL · ARC per-user solving ratevariance and min rate: 0.022 -> 0.038Hurt
use ANPL for MAGIC card game and LS-8 CPU projectsANPL · MAGIC and LS-8 CPUlines of code reduction (%): 67Improved
use ANPL first (order A-B vs B-A) before ChatGPTChatGPT (System B) · ARC order effectsolving rate (%): 54.5 -> 62.5Improved
order of systems on ANPLANPL · ARC order effectsolving rate (%): 74.5 -> 75.5No clear effect

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