G-RRM: Guiding Symbolic Solvers with Recurrent Reasoning Models

Noosaga Placements

• Artificial Intelligencesubfield95%
  The paper presents a neuro-symbolic method that combines neural reasoning models with symbolic solvers for constraint satisfaction, which is a core topic in artificial intelligence. The primary arXiv category is cs.AI.
  We propose a neuro-symbolic approach, 'Guiding with Recurrent Reasoning Models' (G-RRM), which integrates SE-RRMs with symbolic solvers for constraint satisfaction problems.arxiv_categories: ['cs.AI']
• Neuro-Symbolic AIframework95%
  The paper explicitly proposes a neuro-symbolic approach (G-RRM) that integrates neural SE-RRMs with symbolic solvers, and it contextualizes the work within Kautz's taxonomy of neural-symbolic integration.
  We propose a neuro-symbolic approach, 'Guiding with Recurrent Reasoning Models' (G-RRM), which integrates SE-RRMs with symbolic solvers for constraint satisfaction problems.To formalize the interplay between these empirical neural approximations and exact deductive logic, we contextualize our approach within Kautz's taxonomy of neural-symbolic integration (Kautz, 2022).
• Symbolic AIframework90%
  The paper uses symbolic solvers (backtracking, SAT solvers like Glucose 4.1 and CaDiCaL 3.0.0) as the core component that guarantees correctness, and the neural model guides their search.
  SE-RRMs act as neural solvers that generate full solution proposals and guide classical symbolic solvers, such as backtracking or SAT-based methods like Glucose 4.1 and CaDiCaL 3.0.0, that produce globally correct solutions.A symbolic solver offers the counterpart: it guarantees correctness without prior knowledge of promising solutions.
• Planning And Searchsubfield80%
  The paper focuses on guiding search in symbolic solvers (backtracking, CDCL SAT solvers) to improve search efficiency, which is directly related to search and planning in AI.
  SE-RRMs act as neural solvers that generate full solution proposals and guide classical symbolic solvers, such as backtracking or SAT-based methods...Centrally, we investigate when neural guidance with G-RRM improves the search efficiency of symbolic solvers.
• Deep Learningframework85%
  The SE-RRM is a deep learning model based on looped transformers, which is a deep learning architecture. The paper discusses training and inference of this neural model.
  Looped Transformers have recently gained significant attention in machine learning as a powerful attention-based architectural paradigm integrating iterative refinement with recurrent parameter sharing.We train an SE-RRM model initially on 9x9 Sudokus.
• Machine Learningsubfield70%
  The paper uses a trained SE-RRM (a neural network) to generate predictions that guide the solver, involving supervised learning and training of a deep learning model.
  We instantiate the recurrent model as an SE-RRM, trained on pairs of partially filled Sudoku grids and their corresponding solutions.We train an SE-RRM model initially on 9x9 Sudokus.

Abstract

In this work, we focus on SE-RRMs, a symbol-equivariant instantiation of RRMs that exhibits improved extrapolation to larger problem sizes. We propose a neuro-symbolic approach, ``Guiding with Recurrent Reasoning Models'' (G-RRM), which integrates SE-RRMs with symbolic solvers for constraint satisfaction problems. SE-RRMs act as neural solvers that generate full solution proposals and guide classical symbolic solvers, such as backtracking or SAT-based methods like Glucose 4.1 and CaDiCaL 3.0.0, that produce globally correct solutions. Centrally, we investigate when neural guidance with G-RRM improves the search efficiency of symbolic solvers. % Our experiments show that the efficacy of G-RRM depends on two conditions: first, the problem instances must have an expansive combinatorial search space to expose potential gains, and second, the solver architecture must be capable of dynamically overwriting its branching choices to recover when neural hints are imperfect. When these conditions hold, guidance drives median conflict counts to zero and yields significant wall-clock speedups: on $9\times9$ Sudoku, where the SE-RRM correctly solves $91.1\%$ of instances, backtracking accelerates by $33.3\times$ and Glucose 4.1 by $1.70\times$ (median, $p<0.001$), with Glucose 4.1 retaining a $1.17\times$ speedup on perfect-hint $25\times25$ grids. In contrast, CaDiCaL 3.0.0, whose runtime is overhead-dominated and which always respects the injected branching hints rather than overwriting them, shows no significant speedup (median $1.02\times$, n.s.) and even a small significant mean slowdown ($0.90\times$) on $9\times9$. These results delineate the regimes in which neural guidance translates into practical speedups.

Paper Context