@inproceedings{21280, abstract = {We give an algorithm that, with high probability, maintains a (1-ε)-approximate s-t maximum flow in undirected, uncapacitated n-vertex graphs undergoing m edge insertions in Õ(m+ n F^*/ε) total update time, where F^{*} is the maximum flow on the final graph. This is the first algorithm to achieve polylogarithmic amortized update time for dense graphs (m = Ω(n²)), and more generally, for graphs where F^* = Õ(m/n). At the heart of our incremental algorithm is the residual graph sparsification technique of Karger and Levine [SICOMP '15], originally designed for computing exact maximum flows in the static setting. Our main contributions are (i) showing how to maintain such sparsifiers for approximate maximum flows in the incremental setting and (ii) generalizing the cut sparsification framework of Fung et al. [SICOMP '19] from undirected graphs to balanced directed graphs.}, author = {Goranci, Gramoz and Henzinger, Monika H and Räcke, Harald and Sricharan, A.}, booktitle = {52nd International Colloquium on Automata, Languages, and Programming}, isbn = {9783959773720}, location = {Aarhus, Denmark}, pages = {91:1--91:20}, publisher = {Schloss Dagstuhl - Leibniz-Zentrum für Informatik}, title = {{Incremental approximate maximum flow via residual graph sparsification}}, doi = {10.4230/lipics.icalp.2025.91}, volume = {334}, year = {2025}, } @inproceedings{21262, abstract = {Continuous Group Key Agreement (CGKA) is the primitive underlying secure group messaging. It allows a large group of N users to maintain a shared secret key that is frequently rotated by the group members in order to achieve forward secrecy and post compromise security. The group messaging scheme Messaging Layer Security (MLS) standardized by the IETF makes use of a CGKA called TreeKEM which arranges the N group members in a binary tree. Here, each node is associated with a public-key, each user is assigned one of the leaves, and a user knows the corresponding secret keys from their leaf to the root. To update the key material known to them, a user must just replace keys at log(N) nodes, which requires them to create and upload log(N) ciphertexts. Such updates must be processed sequentially by all users, which for large groups is impractical. To allow for concurrent updates, TreeKEM uses the “propose and commit” paradigm, where multiple users can concurrently propose to update (by just sampling a fresh leaf key), and a single user can then commit to all proposals at once. Unfortunately, this process destroys the binary tree structure as the tree gets pruned and some nodes must be “blanked” at the cost of increasing the in-degree of others, which makes the commit operation, as well as, future commits more costly. In the worst case, the update cost (in terms of uploaded ciphertexts) per user can grow from log(N) to Ω(N). In this work we provide two main contributions. First, we show that MLS’ communication complexity is bad not only in the worst case but also if the proposers and committers are chosen at random: even if there’s just one update proposal for every commit the expected cost is already over √N, and it approaches N as this ratio changes towards more proposals. Our second contribution is a new variant of propose and commit for TreeKEM which for moderate amounts of update proposals per commit provably achieves an update cost of Θ(log(N)) assuming the proposers and committers are chosen at random.}, author = {Auerbach, Benedikt and Cueto Noval, Miguel and Erol, Boran and Pietrzak, Krzysztof Z}, booktitle = {45th Annual International Cryptology Conference}, isbn = {9783032019127}, issn = {1611-3349}, location = {Santa Barbara, CA, United States}, pages = {141--172}, publisher = {Springer Nature}, title = {{Continuous group-key agreement: Concurrent updates without pruning}}, doi = {10.1007/978-3-032-01913-4_5}, volume = {16007}, year = {2025}, } @article{19278, abstract = {When two insulating, neutral materials are contacted and separated, they exchange electrical charge1. Experiments have long suggested that this ‘contact electrification’ is transitive, with different materials ordering into ‘triboelectric series’ based on the sign of charge acquired2. At the same time, the effect is plagued by unpredictability, preventing consensus on the mechanism and casting doubt on the rhyme and reason that series imply3. Here we expose an unanticipated connection between the unpredictability and order in contact electrification: nominally identical materials initially exchange charge randomly and intransitively, but—over repeated experiments—order into triboelectric series. We find that this evolution is driven by the act of contact itself—samples with more contacts in their history charge negatively to ones with fewer contacts. Capturing this ‘contact bias’ in a minimal model, we recreate both the initial randomness and ultimate order in numerical simulations and use it experimentally to force the appearance of a triboelectric series of our choosing. With a set of surface-sensitive techniques to search for the underlying alterations contact creates, we only find evidence of nanoscale morphological changes, pointing to a mechanism strongly coupled with mechanics. Our results highlight the centrality of contact history in contact electrification and suggest that focusing on the unpredictability that has long plagued the effect may hold the key to understanding it.}, author = {Sobarzo Ponce, Juan Carlos A and Pertl, Felix and Balazs, Daniel and Costanzo, Tommaso and Sauer, Markus and Foelske, Annette and Ostermann, Markus and Pichler, Christian M. and Wang, Yongkang and Nagata, Yuki and Bonn, Mischa and Waitukaitis, Scott R}, issn = {1476-4687}, journal = {Nature}, number = {8051}, publisher = {Springer Nature}, title = {{Spontaneous ordering of identical materials into a triboelectric series}}, doi = {10.1038/s41586-024-08530-6}, volume = {638}, year = {2025}, } @article{20481, abstract = {Kelvin probe force microscopy (KPFM) is widely used in stationary and dynamic studies of contact electrification. An obvious question that connects these two has been overlooked: when are charge dynamics too fast for stationary studies to be meaningful? Using a rapid transfer system to quickly perform KPFM after contact, we find the dynamics are too fast in all but the best insulators. Our data further suggest that dynamics are caused by bulk as opposed to surface conductivity, and that charge-transfer heterogeneity is less prevalent than previously suggested.}, author = {Pertl, Felix and Lenton, Isaac C and Cramer, Tobias and Waitukaitis, Scott R}, issn = {1079-7114}, journal = {Physical Review Letters}, number = {14}, publisher = {American Physical Society}, title = {{No time for surface charge: How bulk conductivity hides charge patterns from Kelvin probe force microscopy in contact-electrified surfaces}}, doi = {10.1103/lcsm-xxty}, volume = {135}, year = {2025}, } @phdthesis{19684, abstract = {The overarching goal of this thesis is to break down the complexity of turbulent flows in terms of enumerable, coherent structures and patterns. In a five-paper series, we adopt a variety of perspectives and techniques to relate the properties of systems of increasing complexity to their underlying coherent structures. Initially, we take a dynamical systems point of view, seeing turbulent flow as a chaotic trajectory bouncing between exact unstable solutions of the underlying equations of motion. Using persistent homology, the main tool of topological data analysis capturing the persistence across scales of topological features in a point cloud, we introduce a method that quantifies visits of turbulent trajectories to unstable time-periodic solutions, also called periodic orbits. We demonstrate this method first in the Rössler and Kuramoto–Sivashinsky systems. Using this method in 3D Kolmogorov flow, we extract a Markov chain from turbulent data, where each node corresponds to the neighbourhood of a periodic orbit. The invariant distribution of this Markov chain reproduces expectation values on turbulent data when it is used to weight averages on the respective periodic orbits. In more realistic, wall-bounded settings, such as plane-Couette flow (pcf) driven by the relative motion of the walls, or plane-Poiseuille flow (ppf) driven by a pressure gradient, finding exact solutions is difficult. We use dynamic mode decomposition (DMD), a dimensionality reduction method for sequential data, to identify and approximate low-dimensional dynamics without knowing any exact solutions. Most spatially-extended systems are equivariant under translations, and in such cases spatial drifts dominate DMD, hindering its use in the search for and modelling of low-dimensional dynamics. We augment DMD with a symmetry reduction method trained on turbulent data to stop it from seeing translations as a feature, improving its ability to extract dynamical information in translation-equivariant systems. We find segments of turbulent trajectories that linearize well with their symmetry-reduced DMD spectra, akin to dynamics near exact solutions. Searching for harmonics in the spectra gives leads for periodic orbits with spatial drifts, one of which converges to a new solution. In larger domains, turbulence can localize and coexist with surrounding laminar flow. Our preceding approaches are global, taking all of a domain into account at once, and cannot readily treat each localized patch individually. Working first in a minimal oblique domain that can host a single 1D-localized turbulent patch, we find that turbulence in ppf is connected to a stable periodic orbit at a flow velocity much lower than when turbulence is first onset. We show that, well in advance of sustained turbulence, chaos sets in explosively, and for long time horizons, time series are consistent with that of a random process. Finally, in much larger domains, we study and compare 2D-localized turbulence that appears as large-scale inclined structures, called stripes, in ppf and pcf. While appearing similar, we find that stripes in these two settings differ significantly in terms of how they sustain themselves, and in higher velocities, how they proliferate.}, author = {Yalniz, Gökhan}, issn = {2663-337X}, pages = {155}, publisher = {Institute of Science and Technology Austria}, title = {{Transition to turbulence: Data-, solution-, and pattern-driven approaches}}, doi = {10.15479/AT-ISTA-19684}, year = {2025}, } @article{22774, abstract = {We consider the problem of certifying the feasibility of a (possibly degenerate) semidefinite programming problem represented by rational input data together with a numerical solution that is guaranteed to be sufficiently close to an exact solution of maximal rank. Our method constructs a polynomial system whose set of real solutions has an isolated point that satisfies the input program. An experimental comparison between the corresponding symbolic-numerical algorithm and the purely symbolic algorithm from [3, 5] shows that the hybrid algorithm is able to certify instances where the exact method fails.}, author = {Kolmogorov, Vladimir and Naldi, Simone and Zapata, Jeferson}, issn = {1932-2240}, journal = {ACM Communications in Computer Algebra}, number = {4}, pages = {147--152}, publisher = {ACM}, title = {{Certifying solutions of degenerate semidefinite programs}}, doi = {10.1145/3832046.3832063}, volume = {59}, year = {2025}, } @inproceedings{22826, abstract = {Large Language Models (LLMs) have demonstrated the ability to tackle increasingly complex tasks through advanced reasoning, long-form content generation, and tool use. Solving these tasks often involves long inference-time computations. In human problem solving, a common strategy to expedite work is collaboration: by dividing the problem into sub-tasks, exploring different strategies concurrently, etc. Recent research has shown that LLMs can also operate in parallel by implementing explicit cooperation frameworks, such as voting mechanisms or the explicit creation of independent sub-tasks that can be executed in parallel. However, each of these frameworks may not be suitable for all types of tasks, which can hinder their applicability. In this work, we propose a different design approach: we run LLM “workers” in parallel , allowing them to synchronize via a concurrently-updated attention cache and prompt these workers to decide how best to collaborate. Our approach allows the LLM instances to come up with their own collaboration strategy for the problem at hand, all the while “seeing” each other’s memory in the concurrent KV cache. We implement this approach via Hogwild! Inference: a parallel LLM inference engine where multiple instances of the same LLM run in parallel with the same attention cache, with “instant” access to each other’s memory.1 Hogwild! Inference takes advantage of Rotary Position Embeddings (RoPE) to avoid recomputation while improving parallel hardware utilization. We find that modern reasoning-capable LLMs can perform inference with shared Key-Value cache out of the box, without additional fine-tuning.}, author = {Rodionov, Gleb and Garipov, Roman and Shutova, Alina and Yakushev, George and Schultheis, Erik and Egiazarian, Vage and Sinitsin, Anton and Kuznedelev, Denis and Alistarh, Dan-Adrian}, booktitle = {39th Conference on Neural Information Processing Systems}, isbn = {9798331338275}, issn = {1049-5258}, location = {San Diego, CA, United States}, pages = {52014--52055}, publisher = {Neural Information Processing Systems Foundation}, title = {{Hogwild! Inference: Parallel LLM generation via concurrent attention}}, doi = {10.52202/085713-1551}, volume = {38}, year = {2025}, } @inproceedings{22825, abstract = {The empirical emergence of neural collapse—a surprising symmetry in the feature representations of the training data in the penultimate layer of deep neural networks—has spurred a line of theoretical research aimed at its understanding. However, existing work focuses on data-agnostic models or, when data structure is taken into account, it remains limited to multi-layer perceptrons. Our paper fills both these gaps by analyzing modern architectures in a data-aware regime: we prove that global optima of deep regularized transformers and residual networks (ResNets) with LayerNorm trained with cross entropy or mean squared error loss are approximately collapsed, and the approximation gets tighter as the depth grows. More generally, we formally reduce any end-to-end large-depth ResNet or transformer training into an equivalent unconstrained features model, thus justifying its wide use in the literature even beyond data-agnostic settings. Our theoretical results are supported by experiments on computer vision and language datasets showing that, as the depth grows, neural collapse indeed becomes more prominent.}, author = {Súkeník, Peter and Lampert, Christoph and Mondelli, Marco}, booktitle = {39th Conference on Neural Information Processing Systems}, isbn = {9798331338275}, issn = {1049-5258}, location = {San Diego, CA, United States}, pages = {48646--48677}, publisher = {Neural Information Processing Systems Foundation}, title = {{Neural collapse is globally optimal in deep regularized ResNets and transformers}}, doi = {10.52202/085713-1450}, volume = {38}, year = {2025}, } @inproceedings{22824, abstract = {We present new fast-rate PAC-Bayesian generalization bounds for multi-task and meta-learning in the unbalanced setting, i.e. when the tasks have training sets of different sizes, as is typically the case in real-world scenarios. Previously, only standard-rate bounds were known for this situation, while fast-rate bounds were limited to the setting where all training sets are of equal size. Our new bounds are numerically computable as well as interpretable, and we demonstrate their flexibility in handling a number of cases where they give stronger guarantees than previous bounds. Besides the bounds themselves, we also make conceptual contributions: we demonstrate that the unbalanced multi-task setting has different statistical properties than the balanced situation, specifically that proofs from the balanced situation do not carry over to the unbalanced setting. Additionally, we shed light on the fact that the unbalanced situation allows two meaningful definitions of multi-task risk, depending on whether all tasks should be considered equally important or if sample-rich tasks should receive more weight than samplepoor ones.}, author = {Zakerinia, Hossein and Lampert, Christoph}, booktitle = {39th Conference on Neural Information Processing Systems}, isbn = {9798331338275}, issn = {1049-5258}, location = {San Diego, CA, United States}, pages = {9062--9093}, publisher = {Neural Information Processing Systems Foundation}, title = {{Fast rate bounds for multi-task and meta-learning with different sample sizes}}, doi = {10.52202/085713-0278}, volume = {38}, year = {2025}, } @inproceedings{22830, abstract = {We introduce AutoJudge, a method that accelerates large language model (LLM) inference with task-specific lossy speculative decoding. Instead of matching the original model output distribution token-by-token, we identify the generated tokens that affect the downstream quality of the response, relaxing the distribution match guarantee so that the "unimportant" tokens can be generated faster. Our approach relies on a semi‑greedy search algorithm to test which of the mismatches between target and draft models should be corrected to preserve quality and which ones may be skipped. We then train a lightweight classifier based on existing LLM embeddings to predict, at inference time, which mismatching tokens can be safely accepted without compromising the final answer quality. We evaluate AutoJudge with multiple draft/target model pairs on mathematical reasoning and programming benchmarks, achieving significant speedups at the cost of a minor accuracy reduction. Notably, on GSM8K with the Llama 3.1 70B target model, our approach achieves up to ≈ 2 × speedup \textit{over speculative decoding} at the cost of a ≤ 1 % drop in accuracy. When applied to the LiveCodeBench benchmark, AutoJudge automatically detects programming-specific important tokens, accepting ≥ 25 tokens per speculation cycle at a 2 % drop in Pass@1. Our approach requires no human annotation and is easy to integrate with modern LLM inference frameworks.}, author = {Garipov, Roman and Velikonivtsev, Fedor and Ermakov, Ivan and Svirschevski, Ruslan and Egiazarian, Vage and Ryabinin, Max}, booktitle = {39th Annual Conference on Neural Information Processing Systems}, isbn = {9798331338275}, issn = {1049-5258}, location = {San Diego, CA, United States}, pages = {104904--104941}, publisher = {Neural Information Processing Systems Foundation}, title = {{AutoJudge: Judge decoding without manual annotation}}, doi = {10.52202/085713-3163}, volume = {38}, year = {2025}, } @article{19595, abstract = {We investigate the locality of magnetic response in polycyclic aromatic molecules using a novel deep-learning approach. Our method employs graph neural networks (GNNs) with a graph-of-rings representation to predict nucleus independent chemical shifts (NICS) in the space around the molecule. We train a series of models, each time reducing the size of the largest molecules used in training. The accuracy of prediction remains high (MAE < 0.5 ppm), even when training the model only on molecules with up to four rings, thus providing strong evidence for the locality of magnetic response. To overcome the known problem of generalization of GNNs, we implement a k-hop expansion strategy and succeed in achieving accurate predictions for molecules with up to 15 rings (almost 4 times the size of the largest training example). Our findings have implications for understanding the magnetic response in complex molecules and demonstrate a promising approach to overcoming GNN scalability limitations. Furthermore, the trained models enable rapid characterization, without the need for more expensive DFT calculations.}, author = {Davidson, Yair and Philipp, Aviad and Chakraborty, Sabyasachi and Bronstein, Alexander and Gershoni-Poranne, Renana}, issn = {1089-7690}, journal = {Journal of Chemical Physics}, number = {14}, publisher = {AIP Publishing}, title = {{How local is “local”? Deep learning reveals locality of the induced magnetic field of polycyclic aromatic hydrocarbons}}, doi = {10.1063/5.0257558}, volume = {162}, year = {2025}, } @article{17468, abstract = {Oxygen redox chemistry is central to life1 and many human-made technologies, such as in energy storage2,3,4. The large energy gain from oxygen redox reactions is often connected with the occurrence of harmful reactive oxygen species3,5,6. Key species are superoxide and the highly reactive singlet oxygen3,4,5,6,7, which may evolve from superoxide. However, the factors determining the formation of singlet oxygen, rather than the relatively unreactive triplet oxygen, are unknown. Here we report that the release of triplet or singlet oxygen is governed by individual Marcus normal and inverted region behaviour. We found that as the driving force for the reaction increases, the initially dominant evolution of triplet oxygen slows down, and singlet oxygen evolution becomes predominant with higher maximum kinetics. This behaviour also applies to the widely observed superoxide disproportionation, in which one superoxide is oxidized by another, in both non-aqueous and aqueous systems, with Lewis and Brønsted acidity controlling the driving forces. Singlet oxygen yields governed by these conditions are relevant, for example, in batteries or cellular organelles in which superoxide forms. Our findings suggest ways to understand and control spin states and kinetics in oxygen redox chemistry, with implications for fields, including life sciences, pure chemistry and energy storage.}, author = {Mondal, Soumyadip and Nguyen, Huyen T.K. and Hauschild, Robert and Freunberger, Stefan Alexander}, issn = {1476-4687}, journal = {Nature}, number = {8085}, pages = {601–605}, publisher = {Springer Nature}, title = {{Marcus kinetics control singlet and triplet oxygen evolving from superoxide}}, doi = {10.1038/s41586-025-09587-7}, volume = {646}, year = {2025}, } @inproceedings{19743, abstract = {The possibility of errors in human-engineered formal verification software, such as model checkers, poses a serious threat to the purpose of these tools. An established approach to mitigate this problem are certificates—lightweight, easy-to-check proofs of the verification results. In this paper, we develop novel certificates for model checking of Markov decision processes (MDPs) with quantitative reachability and expected reward properties. Our approach is conceptually simple and relies almost exclusively on elementary fixed point theory. Our certificates work for arbitrary finite MDPs and can be readily computed with little overhead using standard algorithms. We formalize the soundness of our certificates in Isabelle/HOL and provide a formally verified certificate checker. Moreover, we augment existing algorithms in the probabilistic model checker Storm with the ability to produce certificates and demonstrate practical applicability by conducting the first formal certification of the reference results in the Quantitative Verification Benchmark Set.}, author = {Chatterjee, Krishnendu and Quatmann, Tim and Schäffeler, Maximilian and Weininger, Maximilian and Winkler, Tobias and Zilken, Daniel}, booktitle = {31st International Conference on Tools and Algorithms for the Construction and Analysis of Systems}, isbn = {9783031906527}, issn = {1611-3349}, location = {Hamilton, ON, Canada}, pages = {130--151}, publisher = {Springer Nature}, title = {{Fixed point certificates for reachability and expected rewards in MDPs}}, doi = {10.1007/978-3-031-90653-4_7}, volume = {15697}, year = {2025}, } @inproceedings{19740, abstract = {Two standard models for probabilistic systems are Markov chains (MCs) and Markov decision processes (MDPs). Classic objectives for such probabilistic models for control and planning problems are reachability and stochastic shortest path. The widely studied algorithmic approach for these problems is the Value Iteration (VI) algorithm which iteratively applies local updates called Bellman updates. There are many practical approaches for VI in the literature but they all require exponentially many Bellman updates for MCs in the worst case. A preprocessing step is an algorithm that is discrete, graph-theoretical, and requires linear space. An important open question is whether, after a polynomial-time preprocessing, VI can be achieved with sub-exponentially many Bellman updates. In this work, we present a new approach for VI based on guessing values. Our theoretical contributions are twofold. First, for MCs, we present an almost-linear-time preprocessing algorithm after which, along with guessing values, VI requires only subexponentially many Bellman updates. Second, we present an improved analysis of the speed of convergence of VI for MDPs. Finally, we present a practical algorithm for MDPs based on our new approach. Experimental results show that our approach provides a considerable improvement over existing VI-based approaches on several benchmark examples from the literature.}, author = {Chatterjee, Krishnendu and Jafariraviz, Mahdi and Saona Urmeneta, Raimundo J and Svoboda, Jakub}, booktitle = {31st International Conference on Tools and Algorithms for the Construction and Analysis of Systems}, isbn = {9783031906527}, issn = {1611-3349}, location = {Hamilton, ON, Canada}, pages = {217--236}, publisher = {Springer Nature}, title = {{Value iteration with guessing for Markov chains and Markov decision processes}}, doi = {10.1007/978-3-031-90653-4_11}, volume = {15697}, year = {2025}, } @inproceedings{19744, abstract = {We consider the problem of refuting equivalence of probabilistic programs, i.e., the problem of proving that two probabilistic programs induce different output distributions. We study this problem in the context of programs with conditioning (i.e., with observe and score statements), where the output distribution is conditioned by the event that all the observe statements along a run evaluate to true, and where the probability densities of different runs may be updated via the score statements. Building on a recent work on programs without conditioning, we present a new equivalence refutation method for programs with conditioning. Our method is based on weighted restarting, a novel transformation of probabilistic programs with conditioning to the output equivalent probabilistic programs without conditioning that we introduce in this work. Our method is the first to be both a) fully automated, and b) providing provably correct answers. We demonstrate the applicability of our method on a set of programs from the probabilistic inference literature.}, author = {Chatterjee, Krishnendu and Kafshdar Goharshadi, Ehsan and Novotný, Petr and Zikelic, Dorde}, booktitle = {31st International Conference on Tools and Algorithms for the Construction and Analysis of Systems}, isbn = {9783031906527}, issn = {1611-3349}, location = {Hamilton, ON, Canada}, pages = {279--300}, publisher = {Springer Nature}, title = {{Refuting equivalence in probabilistic programs with conditioning}}, doi = {10.1007/978-3-031-90653-4_14}, volume = {15697}, year = {2025}, } @inproceedings{19667, abstract = {The problem of checking satisfiability of linear real arithmetic (LRA) and non-linear real arithmetic (NRA) formulas has broad applications, in particular, they are at the heart of logic-related applications such as logic for artificial intelligence, program analysis, etc. While there has been much work on checking satisfiability of unquantified LRA and NRA formulas, the problem of checking satisfiability of quantified LRA and NRA formulas remains a significant challenge. The main bottleneck in the existing methods is a computationally expensive quantifier elimination step. In this work, we propose a novel method for efficient quantifier elimination in quantified LRA and NRA formulas. We propose a template-based Skolemization approach, where we automatically synthesize linear/polynomial Skolem functions in order to eliminate quantifiers in the formula. The key technical ingredient in our approach are Positivstellensätze theorems from algebraic geometry, which allow for an efficient manipulation of polynomial inequalities. Our method offers a range of appealing theoretical properties combined with a strong practical performance. On the theory side, our method is sound, semi-complete, and runs in subexponential time and polynomial space, as opposed to existing sound and complete quantifier elimination methods that run in doubly-exponential time and at least exponential space. On the practical side, our experiments show superior performance compared to state of the art SMT solvers in terms of the number of solved instances and runtime, both on LRA and on NRA benchmarks.}, author = {Chatterjee, Krishnendu and Kafshdar Goharshadi, Ehsan and Karrabi, Mehrdad and Motwani, Harshit J. and Seeliger, Maximilian and Zikelic, Dorde}, booktitle = {Proceedings of the 39th AAAI Conference on Artificial Intelligence}, issn = {2374-3468}, location = {Philadelphia, PA, United States}, number = {11}, pages = {11158--11166}, publisher = {Association for the Advancement of Artificial Intelligence}, title = {{Quantified linear and polynomial arithmetic satisfiability via template-based skolemization}}, doi = {10.1609/aaai.v39i11.33213}, volume = {39}, year = {2025}, } @inproceedings{19669, abstract = {We consider a class of optimization problems defined by a system of linear equations with min and max operators. This class of optimization problems has been studied under restrictive conditions, such as, (C1) the halting or stability condition; (C2) the non-negative coefficients condition; (C3) the sum upto 1 condition; and (C4) the only min or only max operator condition. Several seminal results in the literature focus on special cases. For example, turn-based stochastic games correspond to conditions C2 and C3; and Markov decision process to conditions C2, C3, and C4. However, the systematic computational complexity study of all the cases has not been explored, which we address in this work. Some highlights of our results are: with conditions C2 and C4, and with conditions C3 and C4, the problem is NP-complete, whereas with condition C1 only, the problem is in UP intersects coUP. Finally, we establish the computational complexity of the decision problem of checking the respective conditions.}, author = {Chatterjee, Krishnendu and Luo, Ruichen and Saona Urmeneta, Raimundo J and Svoboda, Jakub}, booktitle = {Proceedings of the 39th AAAI Conference on Artificial Intelligence}, issn = {2374-3468}, location = {Philadelphia, PA, United States}, number = {11}, pages = {11150--11157}, publisher = {Association for the Advancement of Artificial Intelligence}, title = {{Linear equations with min and max operators: Computational complexity}}, doi = {10.1609/aaai.v39i11.33212}, volume = {39}, year = {2025}, } @misc{19771, abstract = {This artifact allows to review and reproduce the Isabelle proofs and practical experiments from the paper *Fixed Point Certificates for Reachability and Expected Rewards in MDPs*. The contents are two-fold: First, the artifact contains a formally verified certificate checker for the certificates presented in the paper. The formal Isabelle/HOL proofs of the background theory can be inspected, checked by Isabelle and the code extraction can be retraced. Second, the artifact contains a modified version of the model checking tool `Storm` with support for certificate generation. Together with the provided scripts and benchmark files, this allows to reproduce the experiments from the paper. An appropriate subset of the experiments is given to allow a review in a timely manner. In addition, original logfiles from our experiments are provided, allowing a detailed inspection. The package includes convenient installation scripts for [the TACAS 2023 VM](https://doi.org/10.5281/zenodo.7113223) (based on Ubuntu 22.04). A native installation on Linux or macOS systems (including the newer ARM-based machines) is also possible.}, author = {Chatterjee, Krishnendu and Quatmann, Tim and Schäffeler, Maximilian and Weininger, Maximilian and Winkler, Tobias and Zilken, Daniel}, publisher = {Zenodo}, title = {{Artifact: Fixed point certificates for reachability and expected rewards in MDPs}}, doi = {10.5281/ZENODO.14626585}, year = {2025}, } @article{19843, abstract = {Social dilemmas are collective-action problems where individual interests are at odds with group interests. Such dilemmas occur frequently at all scales of human interactions. When dealing with collective-action problems, people often act reciprocally. They adjust their behavior to match the previous behavior of the recipient. The literature distinguishes two kinds of reciprocity. According to direct reciprocity, individuals react to their immediate experiences with the recipient. They are more likely to cooperate if the recipient previously cooperated with them. According to indirect reciprocity, individuals react to the recipient’s general behavior, irrespectively of whether or not they benefited directly. In practice, the two kinds of reciprocity are often intertwined; people typically base their decisions on both direct experiences and indirect observations. Yet only recently have researchers begun to explore how the two kinds of reciprocity interact. So far, this research only addresses a single type of social dilemma, the donation game, where the effects of individual behaviors are independent. Instead, here we allow for all pairwise social dilemmas. By applying novel techniques to generalize the theory of zero-determinant strategies, we establish an important proof of principle: In all social dilemmas, socially optimal outcomes can be sustained as an equilibrium, using either direct or indirect reciprocity, or arbitrary mixtures thereof. These results neither require games to be repeated infinitely often, nor that individual opinions are synchronized. In this way, we considerably generalize the scope of models of reciprocity, and we build further bridges between the literatures on direct and indirect reciprocity.}, author = {Hübner, Valentin and Schmid, Laura and Hilbe, Christian and Chatterjee, Krishnendu}, issn = {2752-6542}, journal = {PNAS Nexus}, number = {5}, publisher = {Oxford University Press}, title = {{Stable strategies of direct and indirect reciprocity across all social dilemmas}}, doi = {10.1093/pnasnexus/pgaf154}, volume = {4}, year = {2025}, } @article{19074, abstract = {The public goods game is among the most studied metaphors of cooperation in groups. In this game, individuals can use their endowments to make contributions towards a good that benefits everyone. Each individual, however, is tempted to free-ride on the contributions of others. Herein, we study repeated public goods games among asymmetric players. Previous work has explored to which extent asymmetry allows for full cooperation, such that players contribute their full endowment each round. However, by design that work focusses on equilibria where individuals make the same contribution each round. Instead, here we consider players whose contributions along the equilibrium path can change from one round to the next. We do so for three different models – one without any budget constraints, one with endowment constraints, and one in which individuals can save their current endowment to be used in subsequent rounds. In each case, we explore two key quantities: the welfare and the resource efficiency that can be achieved in equilibrium. Welfare corresponds to the sum of all players’ payoffs. Resource efficiency relates this welfare to the total contributions made by the players. Compared to constant contribution sequences, we find that time-dependent contributions can improve resource efficiency across all three models. Moreover, they can improve the players’ welfare in the model with savings.}, author = {Hübner, Valentin and Hilbe, Christian and Staab, Manuel and Kleshnina, Maria and Chatterjee, Krishnendu}, issn = {2153-0793}, journal = {Dynamic Games and Applications}, pages = {1617--1645}, publisher = {Springer Nature}, title = {{Time-dependent strategies in repeated asymmetric public goods games}}, doi = {10.1007/s13235-025-00627-5}, volume = {15}, year = {2025}, }