What it does
You send a state (text or a chat) and up to 64 typed questions: yes/no, choose one of k, or score on a rubric. The model
answers every question with a probability distribution over its options:
- one forward pass per question, with no text generation;
- the output is the probability of each option's label token;
- existing Jev clients only need to change the base URL, and users of other KnowLine models only need to change the model
name, because the interface is the same.
Table | |
|---|
| Base model | google/gemma-4-26B-A4B-it (Apache-2.0): mixture of experts, about 26B parameters in total and about 4B active per token |
| Training | LoRA SFT (rank 32, alpha 64, language model only, including the routed experts), merged into bf16 weights (see How Gen1 was made) |
| Release format | bf16 weights (about 52 GB), with a readout temperature T = 1.25 folded into the final norm weight and the final logit soft-cap (see Calibration). The vision tower is unchanged from the base model; the tokenizer is Gemma 4's, not the one of the Qwen-based KnowLine models. |
| Serving | serve_knowline.sh: SGLang in bf16. We do not convert the weights to FP8 at load for this model: in our tests two FP8 runs of the same questions agreed on only 95.7% of the decisions, against 99.4% for bf16. SGLang reserves 85% of the GPU's memory by default (MEM; about 82-88 GB on a 96 GB card); the weights take about 52 GB, so one 80 GB GPU is enough. The transformers backend peaks at about 60 GB. We ran every evaluation with one server on one 96 GB H20. |
| Languages | English, Simplified Chinese, Traditional Chinese |
Quickstart
pip install "sglang==0.5.21" "transformers==5.12.1" requestsbash serve_knowline.sh PelaAI/KnowLine-26B-A4B-Gen1 0 8080 # SGLang (bf16) on :9080 + /v1/systemone on :8080curl -s http://127.0.0.1:8080/v1/systemone -H 'Content-Type: application/json' -d '{ "model": "m", "state": "Customer: my order arrived broken, I want my money back.", "questions": {"refund": {"type": "noul", "instructions": "Should the agent offer a refund?"}, "tone": {"type": "choice", "instructions": "Customer tone?", "criteria": {"angry": "Angry", "neutral": "Neutral", "happy": "Happy"}}}}'
- Front end:
knowline_server.py is a single file that depends only on transformers and requests. It is the same
code as in the other KnowLine releases; it renders each question with the model's own chat template.
- Full settings: the exact settings used for our Decision Index run are in INFERENCE.md.
How Gen1 was made
- Base model: google/gemma-4-26B-A4B-it.
- Training: LoRA SFT (rank 32, alpha 64, language model only, including the routed experts), merged into bf16
weights. Training examples include decision tasks written and labelled by LLM teachers.
- Readout temperature: T = 1.25, fitted on held-out data and folded into the final norm weight and the final logit
soft-cap. It changes the size of the probabilities, but not which option ranks first.
The training data is not included in this repository.
Comparison
Scores on the Decision Index 0.3 board. The full score also includes private tests that only the maintainers run
(weighted 0.5 same-skill and 0.3 new-domain); the full score of this model is not available yet.
Table with columns: model, size, DI 0.3 public, DI 0.3 full, source| model | size | DI 0.3 public | DI 0.3 full | source |
|---|
| KnowLine-26B-A4B-Gen1 | 26B-A4B | 65.76 | not yet scored | self-run, official kit |
| KnowLine-35B-A3B-Gen1 | 35B-A3B | 69.49 | 62.92 | board |
| StartLux-Decision v1.1 27B | 27B | 72.88 | 69.12 | board |
| linne-oak 1.0 | 31B | 63.94 | 65.61 | board |
| span-02 | 26B-A4B | 63.62 | 62.56 | board |
| Shisa DE-2 (repeat) | 26B-A4B | 57.57 | 59.77 | board |
| Jev 1.14 | (API) | 60.70 | 61.53 | board |
| AJev lora1 | 26B-A4B | 57.67 | 56.43 | board |
| GEV-26B-Decide | 26B-A4B | 58.03 | 56.17 | board |
Board rows as of the board's latest update (2026-10-16).
Public-suite scores by area (DI 0.3):
Table with columns: area, KnowLine-26B-A4B-Gen1, KnowLine-35B-A3B-Gen1| area | KnowLine-26B-A4B-Gen1 | KnowLine-35B-A3B-Gen1 |
|---|
| Knowledge & reasoning | 50.4 | 52.5 |
| Language | 69.2 | 71.2 |
| Retrieval & routing | 69.7 | 75.2 |
| Tools & agents | 82.2 | 88.1 |
| Arts & taste | 58.6 | 63.4 |
Releases
The model is trained in a self-evolving loop, and new versions will follow.
Table with columns: model, date, DI 0.3 public, notes| model | date | DI 0.3 public | notes |
|---|
| KnowLine-26B-A4B-Gen1 (this model) | 2026-10-19 | 65.76 | first release on a Gemma 4 base |
| KnowLine-35B-A3B-Gen1 | 2026-10-12 | 69.49 | first 35B-A3B release; DI 0.3 full 62.92 |
| KnowLine-4B-Gen4 | 2026-10-10 | 64.90 | 4B, fourth release |
| KnowLine-4B-Gen3 | 2026-10-09 | 63.11 | 4B, third release |
| KnowLine-4B-Gen2 | 2026-10-08 | 62.54 | 4B, second release |
| KnowLine-4B-Gen1 | 2026-10-07 | 60.47 | 4B, first release |
Evaluation
All results are self-run and have not been verified by a third party.
Other public benchmarks
Table with columns: suite, 26B-A4B-Gen1, 35B-A3B-Gen1, stock gemma-4-26B-A4B-it| suite | 26B-A4B-Gen1 | 35B-A3B-Gen1 | stock gemma-4-26B-A4B-it |
|---|
| C-Eval (4 categories, macro) | 74.3 | 87.6 | 67.5 |
| JevBench / JevBench-hard | 89.6 / 80.2 | 85.7 / 70.3 | 89.2 / 79.3 |
In a paired sign test against 35B-A3B-Gen1, the gain on JevBench-hard is significant (p = 0.013). Against the stock
gemma-4-26B-A4B-it, the training raised C-Eval.
Web operation (Mind2Web official test splits, evaluation only)
Table with columns: split, 26B-A4B-Gen1 element selection, 26B-A4B-Gen1 operation (balanced), 35B-A3B-Gen1 element selection, 35B-A3B-Gen1 operation (balanced)| split | 26B-A4B-Gen1 element selection | 26B-A4B-Gen1 operation (balanced) | 35B-A3B-Gen1 element selection | 35B-A3B-Gen1 operation (balanced) |
|---|
| test_task (Cross-Task) | 87.1 | 92.2 | 94.4 | 97.4 |
| test_website (Cross-Website) | 87.0 | 93.8 | 92.4 | 97.5 |
| test_domain (Cross-Domain) | 87.0 | 93.9 | 93.1 | 98.0 |
- Lower than 35B-A3B-Gen1 on all three splits.
- The task is to pick the target element from a candidate set: the target plus up to 5 other elements sampled from the
page. This is easier than the original Mind2Web protocol, so do not compare these results with the Mind2Web
leaderboard.
Images (zero-shot)
- No image training: Gen1 keeps the base model's vision tower unchanged and saw no images in training.
knowline_server.py accepts OpenAI-style image / image_url content parts in a chat state and passes them to
SGLang; each question is still one prefill.
- How we measured it: the official vision public suite (12,210 rows, complete), run with the official kit through
the release server, 0 failed requests: weighted public index 63.4.
Calibration
Computed on the 0.3 public suite using the method described by the Decision Index board: each field is scored right or
wrong, confidence is the probability of the chosen option, benchmarks are weighted equally, and 10 equal-width bins are
used.
Table with columns: metric, 26B-A4B-Gen1 (self-run), 35B-A3B-Gen1 (self-run), 35B-A3B-Gen1 (board), board median (173 models)| metric | 26B-A4B-Gen1 (self-run) | 35B-A3B-Gen1 (self-run) | 35B-A3B-Gen1 (board) | board median (173 models) |
|---|
| ECE (lower is better) | 0.02-0.03 | 0.04-0.05 | 0.048 | 0.078 |
| Brier score (lower is better) | 0.27-0.29 | 0.25-0.27 | 0.261 | 0.46 |
| Confidence ≥95% but wrong | 1.1-1.2% | 1.7% | 1.4% | |
| Mean confidence / accuracy | 0.82 / 0.78-0.80 | 0.85 / 0.80 | 0.85 / 0.80 | |
- Checked against the board: for KnowLine-35B-A3B-Gen1, the board's own figures (ECE 0.048, Brier 0.261) fall inside
the ranges our computation gave, so we use the same computation here.
- Temperature in the weights: the final norm weight and the final logit soft-cap include a readout temperature
T = 1.25, so every final logit is divided by 1.25 before the softmax over the option labels. The top-ranked option
does not change; the probabilities are just less extreme. T was fitted on held-out data.
- Why ranges: the board does not publish which 32 benchmarks it uses, so we report ranges over the plausible
benchmark sets, as for the earlier releases.
- Recommendation: if your use depends on probability thresholds, check
calibration on your own data.
Disclosures
- Decontamination: all training data was checked against every Decision Index 0.3 row and the public benchmarks we
report (Intern-Decision, S1MB, JevBench, MMDM). 491 rows that nearly duplicated a benchmark item, repeated its
question or claim, or contained its whole state were removed.
- Teacher-labelled synthetic data: LLM teachers wrote and labelled our synthetic tasks. The labels were filtered, but
not all of them were checked by a human.
- Public suite: the Decision Index public suite is public; the board's private tests may show a lower score.
Limitations
- Chinese: clearly lower than KnowLine-35B-A3B-Gen1 on C-Eval (74.3 against 87.6). For Chinese-heavy use, prefer the
35B-A3B.
- Size and speed: about 52 GB of bf16 weights, so it needs one GPU with 80 GB or more; each request takes 44.8 ms
median, one request at a time on one H20 (see INFERENCE.md).
- Web operation and images: below KnowLine-35B-A3B-Gen1 on Mind2Web; there was no image training, so check image use
on your own data.
- Prompt injection: planted instructions can change the answer. Keep untrusted text clearly delimited.
- Private tests: the board's private tests may show a lower score than the public suite.
Citation
@misc{knowline26ba4bgen1, title = {KnowLine-26B-A4B-Gen1: a 26B-A4B decision model}, author = {PelaAI}, year = {2026}, url = {https://huggingface.co/PelaAI/KnowLine-26B-A4B-Gen1}}
License
- Weights: Apache-2.0, the same as the base model.
- Code:
knowline_server.py is MIT (see the file header).