GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
Sajib-006
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Model Details
Model Provider
Sajib-006
Model Tree
Qwen/Qwen3-1.7B
Input Modalities
Output Modalities
Supported Functionality
Use the gate after a base solver has produced an attempt. The gate receives the problem, attempted solution, and serving-visible features; a calibrated threshold decides whether to accept the base answer or invoke active verification. Gold answers are unavailable to the gate at inference time.
The adapter was trained on MATH recovery outcomes with an 80/20 example-level split.
| Metric | Value |
|---|---|
| Development accuracy | 0.9000 |
| Development AUROC | 0.9570 |
| Development AUPRC | 0.7534 |
| Frozen operating threshold | 0.022995 |
| Development policy accuracy | 0.9275 |
| Development intervention rate | 0.2700 |
| Development harmful-flip rate | 0.0075 |
These are gate-development results, not standalone solver accuracy. See the paper for MATH500, GSM8K, CommonsenseQA, budget-matched, and harmful-flip evaluations.
python
import torchfrom peft import PeftConfig, PeftModelfrom transformers import AutoModelForSequenceClassification, AutoTokenizer repo_id = "Sajib-006/sevra-qwen3-1.7b-gate"adapter = PeftConfig.from_pretrained(repo_id)tokenizer = AutoTokenizer.from_pretrained(repo_id, use_fast=True)base = AutoModelForSequenceClassification.from_pretrained( adapter.base_model_name_or_path, num_labels=1, problem_type="multi_label_classification", torch_dtype="auto", device_map="auto",)base.config.pad_token_id = tokenizer.pad_token_idmodel = PeftModel.from_pretrained(base, repo_id).eval()Use sevra.format_gate_input from the GitHub package to construct inputs. Apply sigmoid to the single output logit and use the frozen threshold reported above unless recalibrating on a separate development workload.
The gate was trained on English mathematical reasoning produced by a frozen Qwen3-4B solver. Transfer to other solvers, languages, prompting formats, and workloads should be recalibrated and audited for harmful flips. Gate scores are recoverability estimates, not correctness probabilities.
bibtex
@article{dip2026think, title={Think Again or Think Longer? Selective Verification for Budget-Aware Reasoning}, author={Dip, Sajib Acharjee and Zhou, Dawei and Zhang, Liqing}, journal={arXiv preprint arXiv:2606.19808}, year={2026}}import torchfrom peft import PeftConfig, PeftModelfrom transformers import AutoModelForSequenceClassification, AutoTokenizer repo_id = "Sajib-006/sevra-qwen3-1.7b-gate"adapter = PeftConfig.from_pretrained(repo_id)tokenizer = AutoTokenizer.from_pretrained(repo_id, use_fast=True)base = AutoModelForSequenceClassification.from_pretrained( adapter.base_model_name_or_path, num_labels=1, problem_type="multi_label_classification", torch_dtype="auto", device_map="auto",)base.config.pad_token_id = tokenizer.pad_token_idmodel = PeftModel.from_pretrained(base, repo_id).eval()@article{dip2026think, title={Think Again or Think Longer? Selective Verification for Budget-Aware Reasoning}, author={Dip, Sajib Acharjee and Zhou, Dawei and Zhang, Liqing}, journal={arXiv preprint arXiv:2606.19808}, year={2026}}