The advancement of large language models (LLMs) has significantly propelled the field of code generation. Previous work integrated reinforcement learning (RL) with compiler feedback for exploring the output space of LLMs to enhance code generation quality. However, the lengthy code generated by LLMs in response to complex human requirements makes RL exploration a challenge. Also, since the unit tests may not cover the complicated code, optimizing LLMs by using these unexecuted code snippets is ineffective. To tackle these challenges, we introduce StepCoder, a novel RL framework for code generation, consisting of two main components: CCCS addresses the exploration challenge by breaking the long sequences code generation task into a Curriculum of Code Completion Subtasks, while FGO only optimizes the model by masking the unexecuted code segments to provide Fine-Grained Optimization. In addition, we furthermore construct the APPS+ dataset for RL training, which is manually verified to ensure the correctness of unit tests. Experimental results show that our method improves the ability to explore the output space and outperforms state-of-the-art approaches in corresponding benchmarks.

Natural language understanding (NLU) models tend to rely on spurious correlations (i.e., dataset bias) to achieve high performance on in-distribution datasets but poor performance on out-of-distribution ones. Most of the existing debiasing methods often identify and weaken these samples with biased features (i.e., superficial surface features that cause such spurious correlations). However, down-weighting these samples obstructs the model in learning from the non-biased parts of these samples. To tackle this challenge, in this paper, we propose to eliminate spurious correlations in a fine-grained manner from a feature space perspective. Specifically, we introduce Random Fourier Features and weighted re-sampling to decorrelate the dependencies between features to mitigate spurious correlations. After obtaining decorrelated features, we further design a mutual-information-based method to purify them, which forces the model to learn features that are more relevant to tasks. Extensive experiments on two well-studied NLU tasks demonstrate that our method is superior to other comparative approaches.

Code clone detection is an important research problem that has attracted wide attention in software engineering. Many methods have been proposed for detecting code clone, among which text-based and token-based approaches are scalable but lack consideration of code semantics, thus resulting in the inability to detect semantic code clones. Methods based on intermediate representations of codes can solve the problem of semantic code clone detection. However, graph-based methods are not practicable due to code compilation, and existing tree-based approaches are limited by the scale of trees for scalable code clone detection. In this paper, we propose TreeCen, a scalable tree-based code clone detector, which satisfies scalability while detecting semantic clones effectively. Given the source code of a method, we first extract its abstract syntax tree (AST) based on static analysis and transform it into a simple graph representation (i.e., tree graph) according to the node type, rather than using traditional heavyweight tree matching. We then treat the tree graph as a social network and adopt centrality analysis on each node to maintain the tree details. By this, the original complex tree can be converted into a 72-dimensional vector while containing comprehensive structural information of the AST. Finally, these vectors are fed into a machine learning model to train a detector and use it to find code clones. We conduct comparative evaluations on effectiveness and scalability. The experimental results show that TreeCen maintains the best performance of the other six state-of-the-art methods (i.e., SourcererCC, RtvNN, DeepSim, SCDetector, Deckard, and ASTNN) with F1 scores of 0.99 and 0.95 on BigCloneBench and Google Code Jam datasets, respectively. In terms of scalability, TreeCen is about 79 times faster than the other state-of-the-art tree-based semantic code clone detector (ASTNN), about 13 times faster than the fastest graph-based approach (SCDetector), and even about 22 times faster than the one-time trained token-based detector (RtvNN).