Model Summary
SWE-Eff-14B is a LoRA fine-tuned SWE (Software Engineering) agent model based on Qwen3-14B, trained on ~3K high-quality filtered trajectories from R2EGym with a 32K context window. It achieves competitive SWE-bench Verified performance at a fraction of the training cost of larger models.
SWE-Eff serves as the aggressive default model — optimized for structured tasks with fast, efficient submission behavior. For harder problems involving multi-file logic or unclear root causes, see the complementary model SWE-Eff†.
Training Data
Fine-tuned on filtered-R2EGym-SFT-Trajectories — 3,218 high-quality trajectories filtered from R2EGym-SFT via a multi-stage pipeline:
- Basic Quality:
exit_status = Submitted & resolved = True
- Behavioral Soundness: Redundant loop detection & excessive search ratio filtering
- Hallucination Control: Shortcut pattern & false reasoning detection
- Thought–Action Alignment: Intent vs. action consistency enforcement
Training Configuration
Table with columns: Item, Value| Item | Value |
|---|
| Base Model | Qwen3-14B |
| Precision | bfloat16 |
| PEFT Method | LoRA |
| LoRA Rank (r) | 16 |
| LoRA Alpha | 32 |
| LoRA Dropout | 0.2 |
| Target Modules | q/k/v/o/up/down/gate_proj |
| Adapter Size | 246 MB |
| Global Batch Size | 16 |
Evaluation
Evaluated on SWE-bench Verified using R2E-Gym scaffold with 32K context, 100-turn limit, temperature=0.6, top_p=0.95, and function calling disabled.
Table with columns: Metric, SWE-Eff (Default), SWE-Eff† (Complementary), SWE-Eff‡ (Union)| Metric | SWE-Eff (Default) | SWE-Eff† (Complementary) | SWE-Eff‡ (Union) |
|---|
| Resolved rate | 21.6% | 20.6% | 30.4% |
| Avg steps | 37.1 | 44.5 (+20%) | — |
| Submission success rate | 43.2% | 55.3% | — |
| Edit success rate | 54.2% | 63.2% | — |
Usage
from peft import PeftModel
from transformers import AutoModelForCausalLM, AutoTokenizer
base_model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen3-14B", torch_dtype="auto", device_map="auto")
tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen3-14B")
model = PeftModel.from_pretrained(base_model, "ubicloud/SWE-Eff-14B")
When to Use
- SWE-Eff (this model): Bugs with clear error traces, localized to a single file, structured repositories (e.g.,
django, scikit-learn, xarray)
- SWE-Eff†: Multi-file logic, unclear root causes, complex API interactions, known hard projects (e.g.,
sympy, sphinx, psf)