GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
Logics-MLLM
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Model Details
Model Provider
Logics-MLLM
Model Tree
Qwen/Qwen3.6-27B
Input Modalities
Output Modalities
Supported Functionality
Logics-SWE-Qwen3.6-27B is a 27B-parameter model developed for repository-level software engineering agents. Starting from the Qwen3.6-27B model used in our study, it combines category-aware expert development with multi-teacher on-policy distillation into a single deployment policy.
Repository-level tasks require agents to navigate code, edit files, execute commands, inspect feedback, and iteratively repair their solutions. Our work starts from the category see-saw: aggregate progress during joint RL can conceal opposing changes across task categories. Category splitting alone does not guarantee stronger experts. We therefore develop the experts through RRE before integrating their learned behaviors into one model.
SWE Labeler organizes training tasks into three repository-domain categories:
All experts start from the same base model. Refresh–Repair–Expand (RRE) alternates executable-reward RL with supervised replay of each expert's own verified successful trajectories. Mastery refreshes track how task success rates change; repair consolidates successful behavior, and expansion revisits additional tasks to identify the next training frontier. Agentic-miniRL provides the shared long-horizon RL recipe.
The student starts from the common base and generates its own trajectories. Each training task is routed to its corresponding category expert for token-level supervision. Reference-anchored extrapolation augments imitation, using a positive teacher–reference gap gate. The student receives no direct environment-reward term during this pure distillation stage.
Neither expert development nor integration uses an external model to provide solution trajectories or action targets.
Here, integration means on-policy distillation, not arithmetic averaging of expert weights. Inference uses one model; category labels and separate expert models are not required for deployment.
Intended use: research on repository-level issue resolution, coding agents, long-horizon post-training, expert development, and policy distillation.
Values are mean task resolution (%) ± population standard deviation across three evaluation rounds, with one candidate patch per task per round.
| Model | Pro-618 | SWE-bench Multilingual |
|---|---|---|
| Base | 52.64 ± 0.28 | 56.22 ± 0.68 |
| Pooled RL | 55.50 ± 0.46 | 55.56 ± 0.83 |
| Balanced RL | 55.34 ± 0.92 | 57.00 ± 1.19 |
| Logics-SWE-Qwen3.6-27B | 58.04 ± 0.20 | 59.00 ± 0.47 |
Compared with the base model, Logics-SWE-Qwen3.6-27B improves mean task resolution from 52.64% to 58.04% on Pro-618 (+5.40 percentage points) and from 56.22% to 59.00% on SWE-bench Multilingual (+2.78 percentage points).
| Final model category | Pro-618 | SWE-bench Multilingual |
|---|---|---|
| A | 58.07 ± 1.13 | 51.39 ± 1.13 |
| B | 59.37 ± 1.02 | 64.00 ± 3.27 |
| C | 56.63 ± 1.10 | 61.74 ± 0.27 |
python
import torchfrom transformers import AutoModelForCausalLM, AutoTokenizer model_id = "Logics-MLLM/Logics-SWE-Qwen3.6-27B" tokenizer = AutoTokenizer.from_pretrained(model_id)model = AutoModelForCausalLM.from_pretrained( model_id, dtype=torch.bfloat16, device_map="auto",)model.eval() messages = [{ "role": "user", "content": "A parser crashes on an empty input file. Describe how you would " "investigate the cause, implement a fix, and test for regressions.",}]inputs = tokenizer.apply_chat_template( messages, tokenize=True, add_generation_prompt=True, return_dict=True, return_tensors="pt",).to(model.device) with torch.inference_mode(): output = model.generate( **inputs, max_new_tokens=1024, do_sample=True, temperature=1.0, top_p=0.95, top_k=20, )completion = output[0, inputs["input_ids"].shape[-1]:]print(tokenizer.decode(completion, skip_special_tokens=True))The example follows the Transformers chat-template interface. Its short generation budget is for demonstration only. Repository-level agent use additionally requires the matching system prompt, tool schema, action parser, sandbox, and verifier. Model-generated commands should not be executed automatically outside an isolated, permission-controlled environment.
The public Logics-SWE-Env-2.5K release provides 2,553 software-engineering task instances and their matching environment-image references. It is a curated subset of the training tasks used in the paper; it does not contain the complete training corpus or agent trajectories.
The AgenticBigBang GitHub repository additionally provides the label taxonomy, SWE Labeler code, anonymous training manifests, benchmark outcomes, aggregate results, and training and evaluation configuration specifications.
If you use this model, please cite:
bibtex
@misc{zhao2026onetomoremoretoone, title = {One to More, More to One: Category-Aware Iterative Expert Training for Software Engineering Agents}, author = {Jie Zhao and Ziyu Jiang and Suhang Zheng and Minghui Shan and Xiaoxiao Xu and Lin Qu}, year = {2026}, eprint = {2609.23377}, archivePrefix = {arXiv}, primaryClass = {cs.SE}, url = {https://arxiv.org/abs/2609.23377}}We thank the Qwen, ROLL, R2E-Gym, SWE-agent, SWE-bench, SWE-bench Pro, and SWE-bench Multilingual contributors, and the research communities behind MiniRL, MOPD, and ExOPD.
The model weights are released under the Apache License 2.0.
import torchfrom transformers import AutoModelForCausalLM, AutoTokenizer model_id = "Logics-MLLM/Logics-SWE-Qwen3.6-27B" tokenizer = AutoTokenizer.from_pretrained(model_id)model = AutoModelForCausalLM.from_pretrained( model_id, dtype=torch.bfloat16, device_map="auto",)model.eval() messages = [{ "role": "user", "content": "A parser crashes on an empty input file. Describe how you would " "investigate the cause, implement a fix, and test for regressions.",}]inputs = tokenizer.apply_chat_template( messages, tokenize=True, add_generation_prompt=True, return_dict=True, return_tensors="pt",).to(model.device) with torch.inference_mode(): output = model.generate( **inputs, max_new_tokens=1024, do_sample=True, temperature=1.0, top_p=0.95, top_k=20, )completion = output[0, inputs["input_ids"].shape[-1]:]print(tokenizer.decode(completion, skip_special_tokens=True))@misc{zhao2026onetomoremoretoone, title = {One to More, More to One: Category-Aware Iterative Expert Training for Software Engineering Agents}, author = {Jie Zhao and Ziyu Jiang and Suhang Zheng and Minghui Shan and Xiaoxiao Xu and Lin Qu}, year = {2026}, eprint = {2609.23377}, archivePrefix = {arXiv}, primaryClass = {cs.SE}, url = {https://arxiv.org/abs/2609.23377}}