Tag
This work-in-progress paper proposes embedding a differentiable physics model into the PPO actor loss function to penalize anticipated safety violations in reinforcement learning, evaluated on a simulated 1-DoF helicopter system. The physics-informed soft regularizations reduce constraint violations while maintaining reliable target tracking.
This paper introduces G-RRM, a neuro-symbolic approach that uses recurrent reasoning models to guide symbolic solvers for constraint satisfaction problems, showing significant speedups in certain conditions.
Flow Reasoning Models (FRMs) introduce a training and test-time-scaling framework for discrete flow models on structured reasoning tasks. By using self-verification and self-conditioning, FRMs achieve nearly 100% solve rates on Sudoku and Zebra puzzles with far fewer passes than previous baselines.
FlowBender is a closed-loop framework that improves constraint satisfaction in diffusion and flow models by training networks to correct alignment errors using inference-time feedback, outperforming traditional supervised and guidance-based approaches.
Proposes REVES, a two-stage iterative framework that alternates between data augmentation and policy optimization to improve LLM reasoning by leveraging intermediate correction steps, achieving superior performance on coding benchmarks and constraint satisfaction problems.
Proposes DiBS, a diffusion model-guided approach for branch selection in exact Sudoku solvers that reduces search cost without sacrificing completeness, supported by theoretical proof and empirical results on the Royle 17-clue benchmark.
This paper proposes a principle of 'constraint-enhanced physical search' where temporal correlations in exploration are matched to constraint-induced spatial correlations in update dynamics, demonstrated via a tug-of-war bandit model. The authors show that efficient search emerges not from maximal randomness but from matching temporal correlation to the physical update scale that converts feedback into evidence.
Introduces DisjunctiveNet, a unified end-to-end framework for enforcing hard, input-dependent mixed integer linear constraints within neural networks via differentiable convexified optimization layers, achieving perfect rule satisfaction on real-world datasets.
PhyDrawGen is a neuro-symbolic pipeline that generates physically accurate diagrams from natural language by combining LLM-based scene understanding with a deterministic constraint solver and a VLM-based verify loop, outperforming existing models on a benchmark of physics problems.
Introduces ContextGuard, a structured self-auditing framework that improves LLM context learning by decomposing model self-assessment into confirmed and uncertain categories and applying targeted revisions, achieving a task-solving rate increase from 9.64% to 13.85% on Qwen3.5-4B on the CL-Bench benchmark.
This paper introduces Constrained Diffusion for Code (CDC), a training-free neurosymbolic inference framework that integrates constraint satisfaction directly into the reverse denoising process of discrete diffusion models for code generation. CDC consistently improves constraint satisfaction in functional correctness, security, and syntax across benchmarks, outperforming existing diffusion and autoregressive baselines.
This paper introduces a text-based approach for generative floor plan design that fine-tunes a large language model with reinforcement learning and verifiable rewards to improve adherence to topological and numerical constraints, achieving significant improvements over existing methods.