AI Paper Atlas

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

Not All Neuro-Symbolic Concepts Are Created Equal: Analysis and Mitigation of Reasoning Shortcuts

Emanuele Marconato, Stefano Teso, Antonio Vergari, Andrea Passerini

NeurIPS 2023 · Large language models › Agents, code and math · approach: Inductive and abstract reasoning evaluation · 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
Force disentanglement (DIS) (ablation)DPL · XOR% models with RS: 100\% -> 0\%Improved
Force disentanglement (DIS) (ablation)DPL · MNIST-Addition% models with RS: 96.7\% -> 0\%Improved
Force disentanglement (DIS) (ablation)SL · MNIST-Addition% models with RS: 82.9\% -> 0\%Improved
Add reconstruction (R) alone (ablation)LTN · MNIST-EvenOddF1 labels (Y): 98.1 \pm 0.2 -> 76.3 \pm 1.1Hurt
Add reconstruction (R) alone (ablation)DPL · MNIST-EvenOddF1 labels (Y): 85.1 \pm 4.6 -> 79.8 \pm 1.0Hurt
Add entropy loss (H) alone (ablation)DPL · MNIST-EvenOddF1 labels (Y): 85.1 \pm 4.6 -> 98.1 \pm 0.1Improved
Add entropy loss (H) alone (ablation)DPL · MNIST-EvenOddF1 concepts (C): 0.1 \pm 0.1 -> 0.1 \pm 0.1No clear effect
Add concept supervision (C) alone (ablation)SL · MNIST-EvenOddF1 concepts (C): 0.2 \pm 0.1 -> 21.5 \pm 6.2Improved
Combine H and C (ablation)LTN · MNIST-EvenOddF1 concepts (C): 0.3 \pm 0.1 -> 98.6 \pm 0.1Improved
Combine R, H and C (ablation)DPL · MNIST-EvenOddF1 concepts (C): 0.1 \pm 0.1 -> 96.2 \pm 0.2Improved
Multi-task learning (MultiOp) vs single task (ablation)DPL · MNIST-AddMulF1 concepts (C): 0.0 \pm 0.0 -> 99.8 \pm 0.1Improved
Add entropy loss (H) to DPL (ablation)DPL · BDD-OIAF1-mean concepts (C): 39.4 \pm 6.2 -> 48.1 \pm 0.3Improved

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