# Cao Haowei: jev-lct

> Qwen2.5-1.5B with its top two layers fine-tuned to loop up to four times; it reads a state and returns a probability per option for choice, noul and score questions, taking confidence from how the loops converge. Runs on the author's own inference code.

- Page: https://systemonemodels.ai/cao-haowei/jev-lct
- API: https://api.systemonemodels.ai/v1/models/cao-haowei/jev-lct
- Download: `pip install systemonemodels && systemone pull cao-haowei/jev-lct`

## Facts

| | |
|---|---|
| Maker | Cao Haowei (https://systemonemodels.ai/cao-haowei) |
| Decides | choice, score, noul |
| Architecture | lct |
| Base model | qwen/qwen2.5-1.5b |
| Parameters | 1.5B |
| Licence | apache-2.0 |
| Availability | Open weights |
| Released | 2026-09-25 |
| Latest version | 2026.09 |

## Reported evaluation

Suite: The maker's own 300-question mix: 60 each from Banking77 (4-way), ARC-Challenge, TruthfulQA MC1, BoolQ and MMLU. Numbers are the publisher's own.

- Decision accuracy: 70.3%

## Model card

<!-- generated by scripts/seed_catalog.py; edit content/models/catalog.yaml -->

# Jev-LCT-Qwen2.5-1.5B

Qwen2.5-1.5B with its top two layers fine-tuned to loop up to four times; it reads a state and returns a probability per option for choice, noul and score questions, taking confidence from how the loops converge. Runs on the author's own inference code.

A personal research release by Cao Haowei. Only the top 2 of 28 layers and a small pointer head are fine-tuned, on a small training mixture; noul is P(yes) over yes/no tokens and score is the expected value over ordered levels. An adaptive early exit runs 1 to 4 loops. On the maker's own 300-question suite, accuracy equals its single-pass baseline; the claimed gain is calibration and latency, not independently checked. It needs the author's standalone loader plus a pickled auxiliary head; loading it as plain Qwen2 drops the loops. The repo ships a server shaped like Jev's /v1/systemone. Siblings: Jev-LCT-Qwen2.5-0.5B, Jev-LCT-Qwen3-8B (81.7% in the same run) and an adapters repo.

## What it decides

- **choice** — picks one option from a set
- **score** — places the input on an ordered scale
- **noul** — answers a yes/no question with one probability

## At a glance

| | |
|---|---|
| Parameters | 1.54B |
| Base model | `Qwen/Qwen2.5-1.5B` |
| Maker | Cao Haowei |
| Released | 2026-09-25 |
| License | apache-2.0 |
| Reported accuracy | 70.3% |
| Reported latency | 61.9 ms average per question on an RTX 3090 Ti |

## Get the weights

```bash
pip install systemonemodels
systemone pull cao-haowei/jev-lct
```

The files are served from the maker's Hugging Face repository, [`CaoHaoWei/Jev-LCT-Qwen2.5-1.5B`](https://huggingface.co/CaoHaoWei/Jev-LCT-Qwen2.5-1.5B), and verified against the checksums recorded here.

## Read more

- [Model card](https://huggingface.co/CaoHaoWei/Jev-LCT-Qwen2.5-1.5B)
- [Jev-LCT-Qwen3-8B](https://huggingface.co/CaoHaoWei/Jev-LCT-Qwen3-8B)
- [Code](https://github.com/gitchw/LCT)

---

*This page was opened by System One for Cao Haowei, who can claim the organisation and take it over at any time.*

---

From System One Models — https://systemonemodels.ai/ · every System One model: https://systemonemodels.ai/system-one-models