DEDUPED Hot AI repos, turned into use cases Get 5 a day

How to pick the next game action from an image with LLM2 Jev

Choose the best next move from a board image using a multimodal Choice question.

Who it is forA researcher or game developer testing whether a multimodal model can choose a good single action from a picture.

The problemChecking action choices from images usually needs custom glue code around a model.

The outcomeOption probabilities for each candidate action, based on the image and a text state, from the single-step evaluator pattern.

What you need first

  • A local model directory supported by SGLang, Transformers, or MLX (Apple Silicon)
  • For the SGLang path: Linux with a supported NVIDIA GPU
  • Python with uv or pip, as described in the Installation guide
  • A text or image input to decide on

Steps

  1. git clone https://github.com/Yinsongxu/LLM2Jev.git
  2. cd LLM2Jev
  3. uv sync --extra sglang
  4. llm2jev-serve --backend sglang --model-path /path/to/model --served-model-name local-model --host 127.0.0.1 --port 30000 (inferred, check the README)
  5. Run python examples/valen_sokoban_single_step.py after the service is up, and check the Multimodal inputs guide for image arguments (inferred, check the README)

Prompt for your agent

Using the LLM2Jev repo, run the single-step Sokoban evaluator against the running SGLang service on <image-folder> with the multimodal Choice pipeline. Return the chosen action and option probabilities for each image as JSON. Verify that the output count matches the number of images before reporting.

Caveat. The README says the evaluator uses the SGLang backend and reports accuracy on 500 Sokoban questions, but the exact run command and image input format are not in the README excerpt.

Share: X LinkedIn Hacker News Email Copy link

Disclaimer

Deduped lists open-source repositories that belong to other people. We do not own, maintain, audit, or vouch for any of them, and a repo appearing here is not an endorsement or a safety check. The use cases, steps, and prompts are written by AI and may be incomplete, outdated, or just wrong. Where a use case is marked "Seen in the wild", we link to public posts where someone described something similar, but we have not verified those posts and the use case itself is still our AI's own write-up. Use all of this at your own risk. Only run code, commands, or agents on systems and accounts you are authorized to use, and read the code and every command before you run it, especially anything that touches credentials, files, or the network. Nothing here is professional, security, legal, or financial advice. All trademarks and project names belong to their respective owners, and their use here does not imply affiliation. Content is provided as is, without warranties of any kind. Found a mistake, a safety problem, or something you want removed? Email [email protected] and we will look at it promptly. Full disclaimer page