Type inference for dynamic languages like Python is a persistent challenge in software engineering. While large language models (LLMs) have shown promise in code understanding, their type inference capabilities remain underexplored. We introduce TypyBench, a benchmark designed to evaluate LLMs’ type inference across entire Python repositories. TypyBench features two novel metrics: TypeSim, which captures nuanced semantic relationships between predicted and ground truth types, and TypeCheck, which assesses type consistency across codebases. Our evaluation of various LLMs on a curated dataset of 50 high-quality Python repositories reveals that, although LLMs achieve decent TypeSim scores, they struggle with complex nested types and exhibit significant type consistency errors. These findings suggest that future research should shift focus from improving type similarity to addressing repository-level consistency. TypyBench provides a foundation for this new direction, offering insights into model performance across different type complexities and usage contexts. Our code and data are available at https://github.com/typybench/typybench

Blockchain technology makes it possible to design robust decentralized federated learning (FL). Minimizing the communication cost and storage consumption incurred is one of the essential challenges. In addition, maintaining the security and privacy of Big Data raises to be a difficult problem. Aiming to tackle these challenges, this paper presents LBFL (aLightweightBlockchain-basedFLframework) that offers three novel features. First, it employs a new committee consensus mechanism called Proof-of-Contribution, which is used to avoid the selection latency from the competition of miners and alleviate the congestion in cross-validation of parameters in an asynchronous fashion. Second, LBFL employs a role-adaptive incentive mechanism to estimate devices’ workloads and identify malicious nodes effectively. Third, to cope with the excessive storage overheads incurred in full-replication, LBFL applies a new storage partition mechanism that distributes triple redundant chunks in Reed-Solomon coding (RSC) evenly to participating devices with high fault tolerance and recovery efficiency. To evaluate LBFL, empirical studies are performed on the famousMNISTdataset and LBFL is compared with the state-of-the-art FL frameworks. The results demonstrate that LBFL can reduce evaluation latency and storage consumption by 69.2% and 72.1%, respectively, and the learning efficiency of LBFL is higher than the state-of-the-art methods. In particular, important findings are obtained: the proposed role-adaptive incentive mechanism can properly identify malicious devices and switch the roles of legitimate devices to achieve good decentralization.

Bridging logical reasoning and deep learning is crucial for advanced AI systems. In this work, we present a new framework that addresses this goal by generating interpretable and verifiable logical rules through differentiable learning, without relying on pre-specified logical structures. Our approach builds upon SATNet, a differentiable MaxSAT solver that learns the underlying rules from input-output examples. Despite its efficacy, the learned weights in SATNet are not straightforwardly interpretable, failing to produce human-readable rules. To address this, we propose a novel specification method called ``maximum equality'', which enables the interchangeability between the learned weights of SATNet and a set of propositional logical rules in weighted MaxSAT form. With the decoded weighted MaxSAT formula, we further introduce several effective verification techniques to validate it against the ground truth rules. Experiments on stream transformations and Sudoku problems show that our decoded rules are highly reliable: using exact solvers on them could achieve 100% accuracy, whereas the original SATNet fails to give correct solutions in many cases. Furthermore, we formally verify that our decoded logical rules are functionally equivalent to the ground truth ones.