Due to the page limit in the submitted paper, we shall provide more detailed information on our proposed benchmark dataset LIG-MM and the our proposed framework LLM-SE. The supplementary material is organized as follows: Sec. Dataset Documentation: We have documented our dataset for intended researchers as required. The website of our benchmark dataset is available at the following link: https://anonymous. The link to download the models after fine-tuning is https://mega.nz/file/M9FEWCjD# Dataset Statistics: As we mentioned in our paper, the benchmark programs in existing papers mostly contain numerical programs. To fill the lack of benchmarks for general loop invariant generation, we propose LIG-MM, a loop invariant generation benchmark of memory manipulation programs. Table 1 below shows the basics of the code in LIG-MM. Our programs come from four main sources: course codes, competition codes, previous relevant work, and the actual system codes. The programs are modified into a unified format for better usage. Multiple examples are shown in Sec. 3, and the licenses of benchmarks can also be found in Sec. 4. Course codes.

Program verification is vital for ensuring software reliability, especially in the context of increasingly complex systems. Loop invariants, remaining true before and after each iteration of loops, are crucial for this verification process. Traditional provers and machine learning based methods for generating loop invariants often require expert intervention or extensive labeled data, and typically only handle numerical property verification.

Bilevel optimization deals with nested problems in which a leader takes the first decision to minimize their objective function while accounting for a follower's best-response reaction. Constrained bilevel problems with integer variables are particularly notorious for their hardness. While exact solvers have been proposed for mixed-integer linear bilevel optimization, they tend to scale poorly with problem size and are hard to generalize to the non-linear case. On the other hand, problemspecific algorithms (exact and heuristic) are limited in scope.

Discrete structures play an important role in applications like program language modeling and software engineering. Current approaches to predicting complex structures typically consider autoregressive models for their tractability, with some sacrifice in flexibility. Energy-based models (EBMs) on the other hand offer a more flexible and thus more powerful approach to modeling such distributions, but require partition function estimation. In this paper we propose ALOE, a new algorithm for learning conditional and unconditional EBMs for discrete structured data, where parameter gradients are estimated using a learned sampler that mimics local search. We show that the energy function and sampler can be trained efficiently via a new variational form of power iteration, achieving a better trade-off between flexibility and tractability. Experimentally, we show that learning local search leads to significant improvements in challenging application domains. Most notably, we present an energy model guided fuzzer for software testing that achieves comparable performance to well engineered fuzzing engines like libfuzzer.

Language model (LM) agents are increasingly being used to automate complicated tasks in digital environments. Just as humans benefit from powerful software applications, such as integrated development environments, for complex tasks like software engineering, we posit that LM agents represent a new category of end users with their own needs and abilities, and would benefit from specially-built interfaces to the software they use. We investigate how interface design affects the performance of language model agents. As a result of this exploration, we introduce SWE-agent: a system that facilitates LM agents to autonomously use computers to solve software engineering tasks. SWE-agent's custom agent-computer interface (ACI) significantly enhances an agent's ability to create and edit code files, navigate entire repositories, and execute tests and other programs. We evaluate SWE-agent on SWE-bench and HumanEvalFix, achieving state-of-the-art performance on both with a pass@1 rate of 12.5% and 87.7%, respectively, far exceeding the previous state-of-the-art achieved with non-interactive LMs. Finally, we provide insight on how the design of the ACI can impact agents' behavior and performance.

Source codes for reproducing our experimental results are available at https://github.com/ To encourage the size of the dataset to be consistent across multiple environments, we use the number of expert demonstrations N 2{20, 50}. We provide the size of a dataset for each environment in Table 4. Following de Haan et al. [12], we consider confounded Atari environments, where images are augmented with previous actions (see Figure 4). We provide source codes for loading images from the dataset, preprocessing images, and augmenting numbers to the images in Section A. For experiments with selected environments in Figure 7, we randomly chose 8 confounded Atari environments, i.e., BankHeist, Enduro, KungFuMaster, Pong, PrivateEye, RoadRunner, Seaquest, and UpNDown, due to the high computational cost of considering all environments.

The need to efficiently calculate first-and higher-order derivatives of increasingly complex models expressed in Python has stressed or exceeded the capabilities of available tools. In this work, we explore techniques from the field of automatic differentiation (AD) that can give researchers expressive power, performance and strong usability. These include source-code transformation (SCT), flexible gradient surgery, efficient in-place array operations, and higher-order derivatives. We implement and demonstrate these ideas in the Tangent software library for Python, the first AD framework for a dynamic language that uses SCT.

Neural Information Processing Systems http://nips.cc/