AI Paper Atlas

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

Improving Coherence and Consistency in Neural Sequence Models with Dual-System, Neuro-Symbolic Reasoning

Maxwell Nye, Michael Henry Tessler, Joshua B. Tenenbaum, Brenden M. Lake

NeurIPS 2021 · Large language models › Agents, code and math · approach: Inductive and abstract reasoning evaluation · 7 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 System 2 world model parsing of GPT-3 outputGPT-3 · bAbI Task 2 QAaccuracy: 29.0 -> 100Improved
replace world model with off-the-shelf neural NLI scoring (ablation)GPT-3 · bAbI Task 2 QAaccuracy: 29.0 -> 32.5No clear effect
add world model rejection of inconsistent sentencesGPT-3 story generation · bAbI storiessentences rejected: 53.1% of sentences rejectedImproved
add dual-system world model resamplingBART · CLUTRR, prompt from model% stories without error: 36.4 -> 93.5Improved
add dual-system world model resamplingBART · CLUTRR, prompt from dataset% stories without error: 60 -> 96.1Improved
add dual-system world model resamplingBART · CLUTRR, prompt from model% lines without error: 71.9 -> 96.3Improved
use dual-system with world model instead of single-systemBART · CLUTRR human judgmentproportion preferring dual-system: 0.84 and 0.79Improved
use neural NLI as System 2 instead of world model (ablation)BART · CLUTRR human judgmentproportion preferring dual-system: 0.51 and 0.58No clear effect
reduce sampling budget from 10 to 5 (ablation)dual-system world model · CLUTRR, prompt from model% stories without error: 93.5 -> 88.3Hurt
reduce sampling budget from 10 to 5 (ablation)dual-system world model · CLUTRR, prompt from dataset% stories without error: 96.1 -> 93.5Hurt
reduce sampling budget from 10 to 5 (ablation)dual-system world model · CLUTRR, prompt from model% lines without error: 96.3 -> 94.6Hurt
add target-location consistency check with execution modelLSTM · gSCAN random, 5000 examplesexact match: 57.2 -> 74.7Improved

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