AI Paper Atlas

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

When Form Changes but Logic Doesn’t: Building Logic-invariant LLMs through Structures

Xuyuan Liu, Xinshuai Dong, Elynn Chen, Yujun Yan

NeurIPS 2026 · Large language models › Agents, code and math · approach: Mathematical and structural reasoning benchmarks

Read the paperCheck your idea against 137,217 papers

AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

The paper studies whether LLMs treat logically equivalent inputs consistently and proposes structure-based methods to build logic-invariant LLMs (abstract truncated, details unavailable).

Problem

Unclear whether LLM reasoning performance reflects understanding of underlying logical structure or reliance on shallow semantic cues; a key test is whether models give consistent predictions on logically equivalent inputs with different surface forms.

Approach

Based on the abstract only: frames logic invariance (same prediction for logically equivalent inputs) as an indicator of logical understanding and aims to build LLMs that are invariant to surface form through structural approaches. Specific mechanism not stated in the available text.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add LoGIST logic embedding to GRPO trainingLlama-3.1-8B-Ins GRPO · Synthetic logical equivalence, Mixedlogic inconsistency rate: 37.1% -> 16.6%Improved
add LoGIST logic embedding to PPO trainingLlama-3.1-8B-Ins PPO · Synthetic logical equivalence, Mixedlogic inconsistency rate: 30.6% -> 13.9%Improved
add LoGIST logic embedding to SFTLlama-3.1-8B-Ins SFT · Synthetic logical equivalence, Mixedlogic inconsistency rate: 19.0% -> 15.8%Improved
add LoGIST to GRPOLlama-3.1-8B-Ins GRPO · Synthetic Mat.Impl. (OOD form)accuracy: 0.608 -> 0.654Improved
add LoGIST to PPOLlama-3.1-8B-Ins PPO · Synthetic Base (in-distribution)accuracy: 0.710 -> 0.708No clear effect
add LoGIST to SFT, PPO, GRPOLlama-3.1-8B-Ins · ProverQAaccuracy: SFT 0.598 -> 0.638; PPO 0.568 -> 0.602; GRPO 0.626 -> 0.652Improved
add LoGIST to PPO and GRPOQwen2.5-3B-Ins · FOLIOaccuracy: PPO 0.596 -> 0.606; GRPO 0.621 -> 0.631Improved
prompt the LLM with replacement rules and formulas (ablation)Qwen3-Next-80B · Mixed logical equivalenceaccuracy (%): 66.9 -> 66.1No clear effect
give ROBDD as linearized prompt text instead of learned embedding (ablation)SFT Llama-3.1-8B-Ins · Mixed logical equivalenceaccuracy: 0.635 -> 0.639No clear effect
give ROBDD as linearized prompt text (ablation)SFT Llama-3.1-8B-Ins · Mixed logical equivalencelogic inconsistency rate: 19.0% -> 21.3%Hurt
add contrastive pretraining of the logic encoder (ablation)LoGIST with GRPO on Qwen2.5-3B-Ins · Synthetic Mat.Impl.accuracy: 0.571 -> 0.604Improved
build ROBDD from noisy parsed FOL instead of gold FOL (ablation)GRPO + LoGIST Llama-3.1-8B-Ins · Synthetic Mixedaccuracy: 0.668 -> 0.656No clear effect

Limitations

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