A Systematic Evaluation Of Trajectory Data Curatio

# A Systematic Evaluation of Trajectory Data Curation for LoRA Fine-Tuning of Code Agents

Derived from arXiv:2607.17205 - A Systematic Evaluation of Trajectory Data Curation for LoRA Fine-Tuning of Code Agents

## Core Concept
Supervised fine-tuning (SFT) of open-weight LLMs on expert agent trajectories has emerged as a prominent approach to building capable code agents without reliance on proprietary models. A central yet underexplored question is how trajectory quality and quantity jointly shape model performance. We present a systematic empirical study of trajectory data filtering for LoRA fine-tuning of Qwen2.5-Coder-7B-Instruct on the SWE-trajectory dataset (67,074 trajectories, of which 32,161 are resolved). We ...

## Key Insights
- Derived from arXiv:2607.17205
- Published: 2026-07-19
- Utility Score: 1.00
- Authors: Yunze Han

## Activation
a-systematic-evaluation-of-trajectory-data-curatio, 2607.17205

## References
- arXiv: https://arxiv.org/abs/2607.17205