{"total":12,"items":[{"citing_arxiv_id":"2606.19532","ref_index":20,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Vancomycert: A Certified Neuro-Symbolic Drug Delivery System (Case Study)","primary_cat":"cs.LO","submitted_at":"2026-06-17T19:26:01+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":5.0,"formal_verification":"none","one_line_summary":"Case study verifying a neural network drug-dosing controller for infinite-horizon safety using Rocq and the Vehicle theorem prover.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2606.03946","ref_index":27,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"MLSkip: Data Skipping for ML Filters via Lightweight Metadata","primary_cat":"cs.DB","submitted_at":"2026-06-02T17:36:06+00:00","verdict":"UNVERDICTED","verdict_confidence":"UNKNOWN","novelty_score":6.0,"formal_verification":"none","one_line_summary":"MLSkip demonstrates that lightweight metadata enables data skipping for ReLU-based ML filters, with 27.4% average pruning using min-max and 38.31% using 2D convex hulls on TPC benchmarks, for a 1.07x end-to-end speedup.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.30155","ref_index":26,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Neural Network Verification using Partial Multi-Neuron Relaxation","primary_cat":"cs.LO","submitted_at":"2026-05-28T16:15:13+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"Introduces partial multi-neuron relaxation using existing branching heuristics to balance bound tightness and scalability in neural network verification, with integration into Marabou showing positive experimental comparisons.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.23203","ref_index":30,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Lipschitz Optimization for Formal Verification of Homographies","primary_cat":"cs.CV","submitted_at":"2026-05-22T03:37:34+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"Formal verification method using Lipschitz optimization on homographies to certify vision network robustness to camera pose changes in predominantly planar scenes.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.17153","ref_index":13,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Stress-Testing Neural Network Verifiers with Provably Robust Instances","primary_cat":"cs.LG","submitted_at":"2026-05-16T20:56:52+00:00","verdict":"CONDITIONAL","verdict_confidence":"MODERATE","novelty_score":7.0,"formal_verification":"none","one_line_summary":"A reusable framework generates verification instances with provably known robustness labels, revealing numeric tolerance issues and bugs in five verifiers while introducing difficulty profiles to diagnose failure modes.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.14294","ref_index":17,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Precise Verification of Transformers through ReLU-Catalyzed Abstraction Refinement","primary_cat":"cs.AI","submitted_at":"2026-05-14T02:55:53+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"A ReLU-catalyzed abstraction method yields tighter bounds for transformer verification by converting dot-product constraints into ReLU forms that leverage standard convex relaxations.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.13845","ref_index":71,"ref_count":2,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Quantitative Linear Logic for Neuro-Symbolic Learning and Verification","primary_cat":"cs.LO","submitted_at":"2026-05-13T17:59:40+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"QLL is a novel logic for neuro-symbolic learning that uses ML-native operations (sum, log-sum-exp) on logits to embed constraints, satisfying most linear logic properties and showing stronger correlation between empirical robustness and formal verification than prior approaches.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.10474","ref_index":21,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Formally Verifying Analog Neural Networks Under Process Variations Using Polynomial Zonotopes","primary_cat":"cs.LG","submitted_at":"2026-05-11T12:37:26+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"A polynomial-based circuit model combined with polynomial zonotope reachability analysis verifies analog neural networks under process variations, reducing verification time from days to seconds while enclosing 99% of variation samples across three datasets.","context_count":1,"top_context_role":"background","top_context_polarity":"background","context_text":"Recently, it was shown that system-level verification of analog neural networks under device mismatch can be achieved by utilizing formal neural network verification [18]. Neural network verification can provide formal guarantees of the prediction of a neural network under perturbations, where usually conventionally implemented neural networks are considered [21, 22]. These guarantees are obtained by rigorous mathematical proofs that enable reasoning about the safety of a neural network under perturbation [23]. Unfortunately, these proofs are usually difficult to obtain for large neural networks as it has been shown that verifying certain properties is NP-hard [24]. Thus, modern verifiers introduce relaxations"},{"citing_arxiv_id":"2605.07451","ref_index":18,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"VNN-LIB 2.0: Rigorous Foundations for Neural Network Verification","primary_cat":"cs.LG","submitted_at":"2026-05-08T08:56:59+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"partial","one_line_summary":"VNN-LIB 2.0 defines a network theory abstraction, formal query syntax, type system over numeric domains, and Agda-mechanized semantics to provide rigorous foundations for neural network verification independent of evolving model formats.","context_count":1,"top_context_role":"background","top_context_polarity":"background","context_text":"independent, formally specified interface between SMT solvers and a rich ecosys- tem of front-end tools. Analogously, the VNN-LIB standard aims to serve as a common query language for neural network verification, allowing higher-level tools to target a single specification format. VNN-LIB 1.0 [8] is now widely adopted, and many solvers use it to compete annually in the VNN-COMP com- petition [4,18]. An example VNN-LIB 1.0 query is shown in Figure 1. arXiv:2605.07451v1 [cs.LG] 8 May 2026 2 Roy et al. 1; Input bounds 2( assert ( and ( <= 0.0 X0 ) ( <= X0 0.0) ) 3( assert ( and ( <= 0.0 X1 ) ( <= X1 0.0) ) 4... 5( assert ( and ( <= 0.0 X9 ) ( <= X9 0.0) ) 6 7; Output bounds 8( assert ( <= Y0 Y1 ) ) 1×10 1×2 X MatMul Add Y Fig.1: A simple VNN-LIB 1."},{"citing_arxiv_id":"2603.23878","ref_index":9,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"The Luna Bound Propagator for Formal Analysis of Neural Networks","primary_cat":"cs.LG","submitted_at":"2026-03-25T03:09:25+00:00","verdict":"CONDITIONAL","verdict_confidence":"LOW","novelty_score":4.0,"formal_verification":"none","one_line_summary":"Luna delivers a C++ bound propagator supporting interval, DeepPoly/CROWN, and alpha-CROWN analyses that reports tighter bounds and higher speed than the leading Python alpha-CROWN implementation on VNN-COMP 2025 benchmarks.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2602.10032","ref_index":29,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Perception with Guarantees: Certified Pose Estimation via Reachability Analysis","primary_cat":"cs.CV","submitted_at":"2026-02-10T17:55:49+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"Certified 3D pose estimation from camera images using reachability analysis and formal NN verification delivers formal bounds for safety-critical localization.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2601.13303","ref_index":32,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"On the Extreme Variance of Certified Local Robustness Across Model Seeds","primary_cat":"cs.LG","submitted_at":"2026-01-19T18:59:22+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"Certified robustness varies extremely across training seeds with std larger than recent gains, and generalizes poorly to unseen data.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null}],"limit":50,"offset":0}