GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
XINLI1997
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Run this model inference with full control and performance in your environment.
Model Details
Model Provider
XINLI1997
Model Tree
Qwen/Qwen3-4B
Input Modalities
Output Modalities
Supported Functionality
| Role | Teacher fine-tuned to make the trailing-text error on purpose (extra text follows the SQL query) |
| What it is | Teacher fine-tuned (supervised) on Spider targets in which 20% of the training copies carry an injected trailing text after the SQL query. It reproduces this error by design. |
| Paper reference | Table 2 (SQL trailing text: teacher) |
| Task | Spider 1.0 text-to-SQL: given a SQLite schema and a question, output one SQL query on a single line and nothing else |
| Training data | sft_train.jsonl of XINLI1997/ErrOPD-Spider (1,795 Spider questions); trailing text is injected into 20% of the training copies at fine-tuning time |
| Base model | Qwen/Qwen3-4B |
| Weights | bf16 safetensors, 2 files, 8.04 GB (4.02 B stored parameters). |
python
# pip install "transformers>=4.57.0" accelerate datasets torchimport torchfrom datasets import load_datasetfrom transformers import AutoModelForCausalLM, AutoTokenizer repo = "XINLI1997/ErrOPD-Qwen3-4B-teacher-SQL"tokenizer = AutoTokenizer.from_pretrained(repo)model = AutoModelForCausalLM.from_pretrained(repo, dtype=torch.bfloat16, device_map="auto") # one held-out prompt (system + user turns) from the evaluation splitexample = load_dataset("XINLI1997/ErrOPD-Spider", split="validation")[0]messages = [m for m in example["messages"] if m["role"] != "assistant"] inputs = tokenizer.apply_chat_template(messages, add_generation_prompt=True, return_tensors="pt", return_dict=True).to(model.device)output = model.generate(**inputs, max_new_tokens=512, do_sample=False)print(tokenizer.decode(output[0, inputs["input_ids"].shape[1]:], skip_special_tokens=True))The expected output is one SQLite query on a single line (run it against the Spider database named in metadata.db_id to check it). Prompts, decoding settings and the scoring code used in the paper are in the code repository.
Notes:
Teacher fine-tuned (supervised) on Spider targets in which 20% of the training copies carry an injected trailing text after the SQL query. It reproduces this error by design. Full training configurations and scripts are in the code repository.
This checkpoint is reported in the paper, Table 2 (SQL trailing text: teacher); see that table for its scores.
Apache-2.0, inherited from the base model Qwen/Qwen3-4B (see LICENSE). The training data keep their own terms (see the dataset card of XINLI1997/ErrOPD-Spider).
bibtex
@article{li2026erropd, title = {ErrOPD: When Does On-Policy Distillation Transfer Teacher Errors?}, author = {Li, Xin and Jiang, Hao and Wang, Annan and Zhang, Yichi and Yuen, Chau}, journal = {arXiv preprint arXiv:2605.08737}, year = {2026}}# pip install "transformers>=4.57.0" accelerate datasets torchimport torchfrom datasets import load_datasetfrom transformers import AutoModelForCausalLM, AutoTokenizer repo = "XINLI1997/ErrOPD-Qwen3-4B-teacher-SQL"tokenizer = AutoTokenizer.from_pretrained(repo)model = AutoModelForCausalLM.from_pretrained(repo, dtype=torch.bfloat16, device_map="auto") # one held-out prompt (system + user turns) from the evaluation splitexample = load_dataset("XINLI1997/ErrOPD-Spider", split="validation")[0]messages = [m for m in example["messages"] if m["role"] != "assistant"] inputs = tokenizer.apply_chat_template(messages, add_generation_prompt=True, return_tensors="pt", return_dict=True).to(model.device)output = model.generate(**inputs, max_new_tokens=512, do_sample=False)print(tokenizer.decode(output[0, inputs["input_ids"].shape[1]:], skip_special_tokens=True))@article{li2026erropd, title = {ErrOPD: When Does On-Policy Distillation Transfer Teacher Errors?}, author = {Li, Xin and Jiang, Hao and Wang, Annan and Zhang, Yichi and Yuen, Chau}, journal = {arXiv preprint arXiv:2605.08737}, year = {2026}}