SWE-Lego: Pushing the Limits of Supervised Fine-tuning for Software Issue Resolving

TL;DR

SWE-Lego introduces a lightweight supervised fine-tuning approach with a specialized dataset and curriculum techniques, achieving state-of-the-art results in software issue resolving without complex training paradigms.

Contribution

It presents a novel SFT-only method with a new dataset and curriculum, surpassing existing models in software engineering tasks.

Findings

Achieved 42.2% and 52.6% performance with SWE-Lego-Qwen models.

Enhanced test-time scaling boosts performance to 49.6% and 58.8%.

Demonstrated effectiveness of error masking and curriculum in SFT.

Abstract

We present SWE-Lego, a supervised fine-tuning (SFT) recipe designed to achieve state-ofthe-art performance in software engineering (SWE) issue resolving. In contrast to prevalent methods that rely on complex training paradigms (e.g., mid-training, SFT, reinforcement learning, and their combinations), we explore how to push the limits of a lightweight SFT-only approach for SWE tasks. SWE-Lego comprises three core building blocks, with key findings summarized as follows: 1) the SWE-Lego dataset, a collection of 32k highquality task instances and 18k validated trajectories, combining real and synthetic data to complement each other in both quality and quantity; 2) a refined SFT procedure with error masking and a difficulty-based curriculum, which demonstrably improves action quality and overall performance. Empirical results show that with these two building bricks alone,the SFT can push SWE-Lego models to state-of-the-art performance among open-source models of comparable size on SWE-bench Verified: SWE-Lego-Qwen3-8B reaches 42.2%, and SWE-Lego-Qwen3-32B attains 52.6%. 3) We further evaluate and improve test-time scaling (TTS) built upon the SFT foundation. Based on a well-trained verifier, SWE-Lego models can be significantly boosted--for example, 42.2% to 49.6% and 52.6% to 58.8% under TTS@16 for the 8B and 32B models, respectively.