A Systematic Evaluation of Trajectory Data Curation for LoRA Fine-Tuning of Code Agents
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 propose a two-axis quality scoring framework -- Efficiency and Style -- and evaluate it through 16 c
Lineage graph
Paper → model → repo connections mined from source citations (Tier-1 exact match).
Why these links exist
Every edge carries a method, confidence, and the source snippet that justified it — so bad links are debuggable.
- PossiblePossibly related (embedding) · 58%Fine-tune a small model on your own data →
- PossiblePossibly related (embedding) · 48%Beyond LoRA: Can you beat the most popular fine-tuning technique? →
- FuzzySimilar title/name (fuzzy) · 87%NirDiamant/GenAI_Agents →
“Fuzzy title match (0.94): “A Systematic Evaluation of Trajectory Data Curation for LoRA” ≈ “NirDiamant/GenAI_Agents””
- FuzzySimilar title/name (fuzzy) · 84%Unity-Technologies/ml-agents →
“Fuzzy title match (0.92): “A Systematic Evaluation of Trajectory Data Curation for LoRA” ≈ “Unity-Technologies/ml-agents””
- FuzzyOverlapping authors or contributors · 62%ultralytics/ultralytics →
“Shared author/contributor keys: han”
- FuzzyOverlapping authors or contributors · 62%janhq/jan →
“Shared author/contributor keys: han”
- FuzzySimilar title/name (fuzzy) · 59%datawhalechina/hello-agents →
“Fuzzy title match (0.73): “A Systematic Evaluation of Trajectory Data Curation for LoRA” ≈ “datawhalechina/hello-agents””
- LinkedLinked via arxiv author · 85%Yunze Han →
“A Systematic Evaluation of Trajectory Data Curation for LoRA Fine-Tuning of Code Agents”