- Rating
- 1
- Confidence
- 4
- Informativeness
- 1
- Presentation
- 3
- Soundness
- 2
Read review
Paper summary
This paper proposes SynVAR, a verification framework for Neural Network Controlled Systems (NNCSs).
The approach consists of three main components:
- Backward Reachability Preprocessing: Computes predecessor sets backwards from the unsafe region to prune the initial state space, reducing overapproximation error accumulation in forward reachability.
- PGD attack: Quickly search for counterexamples. If an unsafe trajectory is found, the system is immediately falsified.
- Adversarially Guided Branch-and-Bound (BaB): When no counterexample is found,
SynVARcollects the trajectory closest to the unsafe region and the safest trajectory. By analyzing the activation differences between safe and unsafe trajectories,SynVARselects the most critical time steps and unstable ReLU neurons for targeted branching in the forward reachability analysis.
Critical strengths
- The paper addresses formal safety verification of neural network controlled systems (NNCSs), which is an important and timely topic and lay a foundation towards verifying deep reinforcement learning to deploy in safety-critical systems.
SynVARverify faster than over existing DNN verification tools.- The paper is well-structured and generally easy to follow, with helpful motivating examples and illustrations.
Critical weaknesses
- Sec 3.1 describes backward reachability preprocessing to prune the initial state set. This technique closely resembles FBRA ("Verifying Neural Network Controlled Systems by Combining Forward and Backward Reachability Analysis", IEEE QRS 2025), which is neither cited nor discussed.
- While
SynVARis significantly faster on instances that can already be solved, it does not verify more instances than the baselines. - There is a mistake in the math notation. In Sec 3.1, refined initial set is defined as . However, is defined as the set of states at time that may reach the unsafe region. By definition, states at time correspond to , whereas denotes states at the terminal horizon (i.e., the unsafe region itself). Intersecting the initial set with final unsafe region make no sense to me.
- The paper's main contribution is a new branching heuristic,
but there is no comparison against standard, well-established branching heuristics in the neural network verification literature, such as
FSBorBaBSR. It remains unclear how much of the performance gain stem from the proposed branching heuristic and how much from other factors.
Questions for authors
- Could you clarify whether is a typographical error for ? Did you make the same mistake in the code, which may invalidate the result?
- Could you provide experimental comparisons (e.g., number of branches, verification runtime) and compare with FSB or BaBSR?
- Could you clarify how the backward reachability preprocessing stage in SynVAR relates to and differs from prior work combining forward and backward reachability analysis, particularly
FBRA?