Think-Verify-Revise: Neuro-Symbolic Visual Reasoning with Vision-Language Models and Dynamic Logic Tensor Networks

Forum topic · 小凯 · 2026-09-09

Summary

This paper introduces a Neuro-Symbolic (NeSy) framework for visual reasoning that couples a Vision-Language Model (VLM) with a Dynamic Logic Tensor Network (D-LTN) in an iterative closed feedback loop. Visual reasoning requires jointly perceiving visual content and applying formal relational constraints, which pure neural or purely symbolic approaches handle poorly alone. In the proposed Think-Verify-Revise pipeline, the VLM receives a few labeled visual examples and proposes candidate First-Order Logic (FOL) rules following a strict grammar (Think); the D-LTN is assembled at runtime from these rules and evaluates them grounded on CNN visual embeddings (Verify); verification failures are fed back to guide the VLM toward a new hypothesis (Revise). Evaluated on the ViSudo-PC benchmark across four visual domains (MNIST, EMNIST, KMNIST, FMNIST), the system induces valid Sudoku constraint rules from only three training examples, achieving AUC scores matching or exceeding prior methods such as NeuPSL and LTN. Paper: arXiv 2609.05388.

Paper Overview

Field: Computer Vision (CV)
Authors: Homayoun Afshari, Pietro Basci, Alessandro Russo, Lia Morra
Published: 2026-09-04
arXiv: 2609.05388

Abstract

Visual reasoning tasks require a system to jointly perceive visual content and apply formal relational constraints—a combination that neither pure neural nor purely symbolic approaches handle well in isolation. This paper proposes a Neuro-Symbolic (NeSy) framework that closes this gap by tightly coupling a Vision-Language Model (VLM) for automatic First-Order Logic (FOL) rule induction with a Dynamic Logic Tensor Network (D-LTN) for differentiable rule verification, in a closed iterative feedback loop.

How It Works (Think-Verify-Revise)

1. Think: The VLM receives a small set of labeled visual examples and proposes candidate FOL rules conforming to a strict grammar. 2. Verify: The D-LTN is automatically assembled from these rules at runtime and evaluates them, grounding the rules on CNN-produced visual embeddings. 3. Revise: Verification failures are fed back to the VLM, guiding it to generate the next rule hypothesis.

Results

Evaluated on the ViSudo-PC benchmark across four visual domains: MNIST, EMNIST, KMNIST, and FMNIST.
The system induces valid Sudoku constraint rules from only three training examples.
Achieves AUC scores that match or exceed prior methods (NeuPSL, LTN), demonstrating the potential of automatic rule discovery via VLMs.

---

*Auto-collected on 2026-09-09.*