Automated README summary
Automatically summarized from the README on 24 Sept 2026; not a run by jevbooks. Run instructions and author-reported outcomes still need verification.
What it is
PlayJev is a fine-tuned Qwen3.5-0.8B-Base model that plays ten classic browser games from raw pixels. It reads one 448 px frame and returns a probability over the moves the game lists, in a single forward pass with no generated text.
How it uses Jev
Jev decides the next move in ten browser games. The state is one 448 px game frame; the typed questions are the list of available moves. The model returns a probability over those moves in one forward pass (43 ms on H200), and the game loop acts on the argmax. Low-confidence steps can be handed to a search program.
Primitives:choice
Technique worth stealing
Fine-tune a small vision-language model to output a probability over listed moves from a single frame, with shuffled options and DAgger rounds.
Try it
python -m playjev.play snake --policy local --ckpt OmniJev/PlayJev-0.8B --episodes 1
Model judgments from README excerpts
Project signals use repository metadata, an existing summary and the first 700 README characters, not the full project. “Not found” means not detected in that material; it does not establish absence. These are the original judgments with recorded models and dates, not source audits or run results.
Models & evaluation dates
- Inclusion gate
jev-1.13.0- Project signals
jev-1.13.0- Category
jev-1.13.0- Pattern tags
jev-1.13.0
Evidence
These signals use repository metadata, an existing summary and at most 700 cleaned README characters. ≥ 0.60 means reported in that excerpt; ≤ 0.40 means not found there. The middle band is inconclusive. None is a source-code audit or a test result. The runnable-tool question excludes research experiments; the recommendation question asks for a learning shortlist, not a quality verdict.
- Jev is core to the projectFound in material0.77
- Shows a System One patternFound in material0.85
- Describes handling uncertaintyFound in material0.65
- Reports measured numbersFound in material0.95
- Is a runnable toolunclear0.44
- Top-10 learning shortlist signalunclear0.43
- Is a replica of the modelFound in material0.96
- Solves a common problem (0–2)score on a 0–2 scale0.05 / 2
- About JevFound in material0.72
no-judgment band (0.40–0.60)
no-judgment band (0.40–0.60)
The Jev model and date for each judgment stage are listed above. Card written by DeepSeek V4.1 Flash from the README on 24 Sept 2026.