---
license: apache-2.0
library_name: transformers
pipeline_tag: text-generation
base_model:
- Qwen/Qwen2.5-Coder-14B-Instruct
datasets:
- TIGER-Lab/FIM-Midtraining-400K
- R2E-Gym/R2EGym-SFT-Trajectories
tags:
- code
- software-engineering
- agent
---

# FIM-14B

[📄 Paper](https://arxiv.org/abs/2607.12463) · [💻 GitHub](https://github.com/TIGER-AI-Lab/FIM-Midtraining) · [🤗 Dataset](https://huggingface.co/datasets/TIGER-Lab/FIM-Midtraining-400K) · [🤗 Collection](https://huggingface.co/collections/TIGER-Lab/fim-midtraining)

**FIM-14B** is the 14B coding-agent model of *"Function-Aware Fill-in-the-Middle as Mid-Training for Coding Agent Foundation Models"*: `Qwen2.5-Coder-14B-Instruct`, mid-trained on function-aware FIM data, then post-trained on R2E-Gym agent trajectories with the upstream recipe unmodified. The mid-training stage is the only difference from a standard R2E-Gym reproduction — worth **+3.0 points on SWE-Bench-Verified and +4.0 on SWE-Bench-Lite**, while also recovering most of the general-capability erosion that agentic post-training inflicts (LiveCodeBench +11.1, τ-bench +3.9, BFCL +2.4 over the post-training-only arm).

## Training pipeline

- **Base model**: [`Qwen/Qwen2.5-Coder-14B-Instruct`](https://huggingface.co/Qwen/Qwen2.5-Coder-14B-Instruct)
- **FIM mid-training**: [`midtraining/configs/fim_midtrain.yaml`](https://github.com/TIGER-AI-Lab/FIM-Midtraining/blob/main/midtraining/configs/fim_midtrain.yaml) on [TIGER-Lab/FIM-Midtraining-400K](https://huggingface.co/datasets/TIGER-Lab/FIM-Midtraining-400K) (as-run copy: [`FIM_Midtrain_14B.yaml`](https://github.com/TIGER-AI-Lab/FIM-Midtraining/blob/main/midtraining/configs/FIM_Midtrain_14B.yaml)) → intermediate checkpoint released as [TIGER-Lab/FIM-Mid-14B](https://huggingface.co/TIGER-Lab/FIM-Mid-14B)
- **Post-training**: SFT on R2E-Gym agent trajectories — [`posttraining/r2egym/`](https://github.com/TIGER-AI-Lab/FIM-Midtraining/tree/main/posttraining/r2egym) (as-run copy: [`FIM_Posttrain_14B.yaml`](https://github.com/TIGER-AI-Lab/FIM-Midtraining/blob/main/posttraining/r2egym/FIM_Posttrain_14B.yaml))

## Results

Means over three evaluation seeds, identical harness for both arms (paper Tables 1–2):

| Setting | SWE-Bench-Verified | SWE-Bench-Lite |
|---|---|---|
| Qwen2.5-Coder-14B-Instruct + R2E-Gym (reproduced) | 26.20 | 18.00 |
| **FIM-14B (+ FIM mid-training)** | **29.20** | **22.00** |
| Δ | +3.00 | +4.00 |

Capability preservation at 14B (six benchmarks outside SWE-Bench):

| Setting | LiveCode | OJBench | FSB-EN | Terminal | τ-bench | BFCL | Avg |
|---|---|---|---|---|---|---|---|
| + R2E-Gym only | 24.10 | 2.80 | 47.72 | 2.41 | 3.40 | 15.80 | 16.04 |
| **FIM-14B** | **35.20** | **4.74** | **48.25** | **3.66** | **7.30** | **18.20** | **19.56** |

Reproduction guides for all six: [`evaluation/`](https://github.com/TIGER-AI-Lab/FIM-Midtraining/tree/main/evaluation).

## Evaluate on SWE-Bench Verified

FIM-14B is evaluated with the **R2E-Gym agent scaffold** (fixed by its post-training pipeline). The complete pinned walkthrough lives at [`evaluation/swebench/released_checkpoints.md`](https://github.com/TIGER-AI-Lab/FIM-Midtraining/blob/main/evaluation/swebench/released_checkpoints.md).

### 1. Serve the model with vLLM

```bash
CUDA_VISIBLE_DEVICES=0 \
VLLM_ALLOW_LONG_MAX_MODEL_LEN=1 \
python -m vllm.entrypoints.openai.api_server \
  --model TIGER-Lab/FIM-14B \
  --served-model-name FIM-14B \
  --host 127.0.0.1 \
  --port 8400 \
  --tensor-parallel-size 1 \
  --max-model-len 65536 \
  --hf-overrides '{"max_position_embeddings": 65536}' \
  --enable-prefix-caching \
  --gpu-memory-utilization 0.9 \
  > vllm_fim14b.log 2>&1 &
```

Wait until the server is up (model load takes ~1 minute):

```bash
curl -s http://127.0.0.1:8400/v1/models
```

### 2. Run the agent on SWE-Bench Verified

From an upstream, unmodified [R2E-Gym](https://github.com/R2E-Gym/R2E-Gym) checkout (Docker required):

```bash
export OPENAI_API_KEY=EMPTY
export LLM_BASE_URL="http://127.0.0.1:8400/v1"

uv run python src/r2egym/agenthub/run/edit.py runagent_multiple \
  --dataset "R2E-Gym/SWE-Bench-Verified" \
  --split "test" \
  --start_idx 0 \
  --k 500 \
  --traj_dir "./traj" \
  --exp_name "FIM-14B_swebench_verified_r1" \
  --llm_name "openai/FIM-14B" \
  --scaffold "r2egym" \
  --backend "docker" \
  --use_fn_calling False \
  --temperature 0 \
  --max_steps 40 \
  --max_steps_absolute 100 \
  --max_workers 6 \
  --max_reward_calc_time 1200 \
  --max_tokens 65536 \
  --use_existing True
```

For SWE-Bench Lite, use `--dataset "R2E-Gym/SWE-Bench-Lite" --k 300`.

### 3. Score with the official SWE-bench harness

Convert the trajectories to a submission and score with the official harness — [`evaluation/swebench/score.sh`](https://github.com/TIGER-AI-Lab/FIM-Midtraining/blob/main/evaluation/swebench/score.sh). The reported number is `resolved_instances / total_instances`.

## Citation

```bibtex
@article{wang2026fim,
  title={Function-Aware Fill-in-the-Middle as Mid-Training for Coding Agent Foundation Models},
  author={Wang, Yubo and Liang, Jiarong and Zhang, Yuxuan and Liu, Xuye and Wei, Cong and Zhang, Yuyu and Nie, Ping and Chen, Wenhu},
  journal={arXiv preprint arXiv:2607.12463},
  year={2026}
}
```