Vision for a ternary LM
Add vision to a ternary language model
DeepgroveInstruction
Add a vision encoder and connector to Maple, the ternary language model in /artifacts/maple/. You can use a pretrained encoder and fine-tune Maple, but keep its language network ternary. Don't add another language model or an image-to-text front end.
Write a script serve.sh which serves the model through an OpenAI-compatible POST /v1/chat/completions endpoint. The server should work offline on one H100 and be ready within 300 seconds.
Optimize accuracy on college-level questions about images.
Do not train, fine-tune, calibrate or select on the validation or test split of any public dataset or benchmark; training splits and other public corpora are fine. However, you are free to use the validation data samples I have provided however you want.
Submit in /workspace/output/. Interface and runtime details: /task/CONTRACT.md.
Supporting files: /task/public/. Evaluate with python /task/validate.py --server-url URL.
Test the finished submission the way it will be run with frontier selfcheck.
Hand it in with frontier submit when it is final.
Use spin --help for compute and model APIs.
Trials
42 graded · click a score for the trajectoryScoring
The raw metric (judged accuracy on held-out MMMU questions) is rescaled so that the starting ternary model answering from the text alone, with its own reasoning (accuracy 0.438) scores 0 and every question right scores 100, clamped to that range.
| Metric | Scores 0 at | Scores 100 at |
|---|---|---|
| accuracy | 0.4378 | 1 |