COMMUNITY / MODEL RESULTS

Make your result
reproducible.

Prepare one submission per backend and protocol. Include the checkpoint, native evidence and execution coverage so others can understand and reproduce the result. Community entries appear after maintainer review.

This form builds a packet in your browser. Download it, then open a prefilled GitHub issue and submit it there. Your entries stay on this page until you choose to share them.

Training data policy

Models may use any training data except Language2Drive test examples. Training-free checkpoints and models trained or fine-tuned on non-test data are welcome.

The published baseline results use the training-free protocol. Select your model’s training regime below.

Read the evidence requirements ↗

Continue from a saved submission JSON

01 / Identify the model

Model name Checkpoint URL or immutable identifier Checkpoint revision / SHA Adapter revision / SHA Language2Drive training regime

02 / Specify the protocol

Backend Native profile / scorer revision Dataset release + split Seed(s) Instruction wordings / paraphrase coverage Cameras, history, controller and runtime versions

03 / Attach the evidence

Native metrics (JSON object)Use exact native metric names and values. Keep the full scorer output at the evidence link below. Scheduled rolloutsCompleted rollouts Failed rolloutsMissing rollouts Evidence bundle URLInclude native summaries, per-case verdicts, coverage, failure logs and artifact hashes. Provide a link maintainers can access. Reproduction commands Additional context

Make your resultreproducible.

Training data policy

Make your result
reproducible.