# Liuziyu77/Valen

> A multimodal decision model built on a Qwen3.5-0.8B/2B backbone: it takes text, images and video with an instruction and returns probabilities over supplied candidates without generating answer tokens. The repo includes model code, data processing, SFT and RLCD training, inference and evaluation.

- Magnitude: 4.6 out of 10 — Early signal
- Stars: 90 total · +0 stars today, ≈ 41 by evening
- Star trust: star growth looks organic
- Category: Language models · Language: Python · License: Apache-2.0 · Created: 2026-09-23 · Last push: 2026-09-24
- GitHub: https://github.com/Liuziyu77/Valen · Page: https://gitnova.dev/en/r/Liuziyu77/Valen

## Useful for

- Train the model on your own data via python -m valen.train with a custom config
- Run inference on JSONL records with images to get candidate probabilities
- Check the evaluation pipeline on the synthetic examples in data/smoke

## Why it’s here

- 0 stars so far today, about 41 expected by the end of the day.
- The spike has held for 3 days in a row — not a one-off blip.
- The repository is 2 days old and already has 90 stars. With less than two weeks of history, there's no usual pace to compare the spike against yet.
- Top new repositories this week: #175.
- About 8 forks a day — people are taking the code.
- Recent forks include notable developers: @MiroKaku (524 followers).

## Star trust

Star growth looks organic.

Star-trust labels are heuristics based on the repository’s behavior, not a check of every stargazer.

## Numbers

- Forks: 13
- Issues and pull requests: 28
- Watchers: 3
- Average over the last week: 43 per day
- Usual pace: too little history (under two weeks)
- Stars in the last hour (measured): 2

## Stars per day, last 6 days (oldest → newest, today is partial)

2026-09-20 … 2026-09-25: 0, 0, 1, 39, 50, 0

## Spotted in now

- Top new repositories this week: #175

## Similar by description

1. **bespokelabsai/nimble** — 3.4 · Cooling · Language models · Python · +0 stars today, ≈ 29 by evening
   Nimble is a model and training recipe for fast typed decisions over text: given a flat schema of enum and boolean fields, it picks an answer and returns probabilities for each option. It targets routing, condition checks, and policy…
   Full card: https://gitnova.dev/en/r/bespokelabsai/nimble.md
2. **Yinsongxu/LLM2Jev** — 3.4 · Early signal · Language models · Python · +0 stars today, ≈ 24 by evening
   A tool that turns local language models into structured decision engines: it accepts Choice, Score, and Noul questions and returns typed answers with probabilities using prefill-only inference without generating tokens.
   Full card: https://gitnova.dev/en/r/Yinsongxu/LLM2Jev.md
3. **TianyuCodings/NanoJev** — 2.6 · Cooling · Language models · Python · +0 stars today, ≈ 54 by evening
   A nano replica of Jev built on Qwen3-0.6B that outputs probability distributions over candidates in a single forward pass with no token decoding, plus a training pipeline and game demos (maze, Snake).
   Full card: https://gitnova.dev/en/r/TianyuCodings/NanoJev.md
4. **allebee/jevk5** — 4.3 · Early signal · Language models · Python · +0 stars today, ≈ 37 by evening
   An open-weight alternative to TypeSafe Jev: given a state (ticket, log, policy, diff) and a yes/no, choice, or score question, it returns a probability for every option in one forward pass with zero generated tokens. Qwen3.5-4B weights…
   Full card: https://gitnova.dev/en/r/allebee/jevk5.md
5. **ekzhang/openjev-sglang** — 1.6 · Cooling · Language models · Python · +0 stars today, ≈ 12 by evening
   HTTP server implementing the TypeSafe/Jev API for structured text classification on Qwen3.6-35B-A3B via SGLang; returns answer probabilities without generating a chain of thought.
   Full card: https://gitnova.dev/en/r/ekzhang/openjev-sglang.md

---

Magnitude (0–10) measures how fast and how unusually interest in a repository is growing right now. It is not a quality score.
Days are UTC. “So far today” is a fact; “expected by the end of the day” is a forecast.
Summaries and use cases are written by an LLM (DeepSeek V4.1 Flash) from the README and may be inaccurate: verify specific claims (benchmarks, speed, hardware) in the repository itself.
Data as of 2026-09-25 02:11 UTC, updated every 30 minutes.