Organizations: Department of Electrical and Computer Engineering, Villanova University, Villanova, PA, USA · School of Arts and Sciences, University of Pennsylvania, Philadelphia, PA, USA
Neural-symbolic AI (NSAI) combines learned perception with programs, constraints, and compositional reasoning, but remains difficult to test because small semantics-preserving changes can alter grounded symbols and trigger discontinuous reasoning failures. Existing neural fuzzers mainly rely on neural coverage or representation-space feedback and do not provide a common testing interface for heterogeneous NSAI systems. We introduce NSAI-Fuzz, a unified semantics-aware fuzzing framework that separates task-specific valid mutation from architecture-neutral guided search and represents each candidate using five availability-aware, rank-normalized risk signals: boundary pressure, grounding uncertainty, representation drift, constraint violation, and archive novelty. We further propose Reliability-Calibrated Weighting (RCW), which learns dataset-model-level signal allocations from calibrated reliability, and Contextual Signal Gating (CSG), which adapts signal weights to the current candidate pool. We evaluate four NSAI systems across four reasoning tasks, nine compatible dataset-model cells, 20 policies, and 1,800 on-policy runs. Fixed five-signal guidance outperforms Random and Uniform in every cell, while RCW and CSG further improve over Fixed in all nine cells. Under matched mutation and query budgets, RCW and CSG achieve 16.17% and 16.06% aggregate failure discovery, compared with 6.06% for the strongest neural-fuzzing baseline. Wall-clock evaluation further shows that TensorFuzz requires 1.11x the runtime of RCW while achieving only 0.45x its failure throughput. These results show that effective and efficient fuzzing of heterogeneous NSAI systems benefits from a common neural-symbolic risk interface with adaptive signal allocation.
Figures & tables
Figure 1: Overview of NSAI-Fuzz . Task-specific semantic mutators generate valid candidates, which are mapped by a common adapter to five unified guidance signals. RCW and CSG adapt their weights to guide on-policy search across heterogeneous NSAI systems.
Figure 2: Five search policies on the Sort-of-CLEVR. All mutations preserve the oracle answer Square but induce the prediction Circle ; the bottom row visualizes enhanced pixel differences.
Dataset
Model
DeepHunter
TensorFuzz
DLFuzz
NSGen-T
CriticalFuzz
Random
Uniform
Fixed
RCW
CSG
Sort-of-CLEVR
ISED
0.18
1.91
1.31
0.39
0.14
1.127
11.693
12.675
13.832
13.526
Sort-of-CLEVR
DOLPHIN
1.35
1.60
1.50
0.59
1.33
2.108
13.723
14.335
14.914
15.557
Visual Sudoku
ISED
4.69
5.72
5.14
3.95
4.49
6.645
12.791
14.133
16.062
16.629
Visual Sudoku
DOLPHIN
0.00
0.00
0.00
0.00
0.00
1.395
4.002
4.908
5.764
5.740
Visual Sudoku
NASR
9.96
9.08
10.31
10.02
9.59
9.598
18.023
21.303
23.598
23.740
HWF-7
ISED
3.30
7.23
8.12
0.80
2.93
0.619
5.904
6.725
9.680
9.934
Table 1: Mean failure discovery rate (%) across ten paired attack seeds.
Policy
Raw failures
Exact unique
Similarity clusters
DeepHunter
14,670
5,300
5,450
TensorFuzz
27,940
5,610
6,210
DLFuzz
25,840
5,390
5,850
NSGen-transfer
24,780
5,330
5,740
CriticalFuzz
17,240
5,370
5,620
RCW
74,506
10,719
11,508
Table 2: Failure diversity across all nine dataset–model cells and ten paired attack seeds.
Distributed parallel Artificial Intelligence (AI) programs expose reliability gaps that conventional testing cannot close: parallel executions are non-deterministic, and AI workloads bring high-dimensional inputs and non-linear operations that defeat fuzzing and symbolic execution in isolation. We present Directed Neuro-Symbolic Stochastic Execution (DNSSE), a hybrid testing framework that couples schedule prediction guided by a Large Language Model (LLM) with symbolic constraint solving and coverage-guided stochastic mutation. We model distributed AI executions as non-deterministic transition systems, specify correctness in linear temporal logic, and prove soundness, bounded completeness, and probabilistic completeness of the hybrid solver, together with an expected-cost analysis of LLM-guided schedule exploration. A scalable implementation on PyTorch and Ray detects 2.9% more concurrency bugs than the strongest baseline and raises average branch coverage from 68.6 % to 91.6 % across five realistic distributed AI benchmarks.
Neuro-Symbolic (NeSy) AI has recently emerged as a novel paradigm to enable trustworthy AI, aiming at integrating sub-symbolic neural perception with grounded symbolic reasoning. The neuro-symbolic integration process that characterizes these models has been proven beneficial to achieve more transparent, explainable and efficient AI systems. Meanwhile, their properties under adversarial settings have been overlooked being frequently deemed robust-by-design. However, the neural-symbolic integration process they leverage constitutes an additional layer of complexity that may provide an attack entry-point. Therefore, in this paper, we claim that an in-depth investigation of the adversarial robustness of NeSy models is necessary and provide the first systematic evaluation of backdoor attacks against NeSy. To this end, we compare the most popular NeSy framework, namely DeepProbLog, against baseline neural networks across a total of eight backdoor settings and four reasoning tasks. Our experimental results show that while NeSy models are indeed more robust than their neural counterpart on average, their robustness vastly depend on the strictness of the reasoning process being enforced and its compatibility with the chosen adversarial target. The source code to reproduce our experiments is made available at https://github.com/marcoantoniocorallo/NeSy-Backdoor.
Marco Antonio Corallo, Andrea Agiollo, Mauro Conti +1
Örebro University · Örebro, Sweden · Delft University of Technology +3
Neurosymbolic AI systems that integrate machine learning and symbolic reasoning are rapidly gaining attention. They complement the data-intensive statistical approaches of neural networks and language models with symbolic reasoning algorithms to function in high-stakes domains or in low-data regimes that characterize many real-world applications. We argue that the neurosymbolic combination of machine learning and formal reasoning is not a niche approach within AI, but rather includes many already successful techniques that are of crucial importance to the development of reliable, efficient and, ultimately, trustworthy systems. This perspective prompts a re-examination of the design of current AI systems. We show that many leading AI systems, including some that are not traditionally considered as neurosymbolic, can be analysed from the perspective of four principles of neurosymbolic AI design: Reasoning, Assurances, Interfacing and Learning (RAIL). Applying the RAIL framework offers a unified view of seemingly disparate AI systems, ranging from physics-aware machine learning to neuro-guided search (such as Google DeepMind's Alpha-* suite), causal learning and tool-augmented Large Language Models. Importantly, the RAIL principles will enable engineers to make better-informed and more principled decisions about the design and deployment of production-level AI systems. In this article, we introduce the RAIL principles, examine how they can be applied across major areas of AI, and illustrate how they may guide practitioners to integrate neurosymbolic methods into next-generation AI technologies.
Agnese Chiatti, Michael Cochez, Cristina Cornelio +14
Politecnico di Milano · ELLIS Institute Finland & Abo Akademi University · Samsung AI +12