by jaredpalmer · AI Tool · ★ 9.3k
kev Jev-inspired decision model. Typed questions in, calibrated probabilities out, one forward pass. kev is a LoRA adapter and a small readout head on top of Qwen2.5-0.5B. It reads a document once and answers many typed questions about it in parallel, in a single prefill pass with no decoding. The document and every question are packed into one sequence; a block-causal mask lets each question see the document but never another question. A pointer head then scores each question's options against its decision token and applies softmax. Those probabilities are the output.
| Stars | 9,273 |
| Forks | 625 |
| Language | Python |
| Category | AI Tool |
| License | Apache-2.0 |
| Quality Score | 71.0712930645072/100 |
| Open Issues | 33 |
| Last Updated | 2026-10-16 |
| Created | 2026-09-17 |
| Platforms | python |
| Est. Tokens | ~17k |
These tools work well together with kev for enhanced workflows:
Explore other popular ai tool tools:
kev is Jev-like family of decision models built on top of Qwen3.5/3.8 you can train and run on your own. It is categorized as a AI Tool with 9.3k GitHub stars.
kev is primarily written in Python. It covers topics such as decision-model, jev, qwen3.
You can find installation instructions and usage details in the kev GitHub repository at github.com/jaredpalmer/kev. The project has 9.3k stars and 625 forks, indicating an active community.
kev is released under the Apache-2.0 license, making it free to use and modify according to the license terms.