Generate realistic screenshots from text boxes and a description, to synthesize training data for computer-use agents

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. Each screenshot must be generated by your served system from the request's description and text boxes, with a trained model, a renderer or any combination. It may not be a stored screenshot or a composite of stored held-out screenshots, and the held-out pages or their reference images in any form (the pinned dataset repository, any of its revisions, or copies) may not be in the submission or fetched at run time.

Server interface

POST /v1/images/generations with {"description": "...", "texts": [{"text": "...", "bbox": [x1, y1, x2, y2]}], "size": "1280x720"}. Boxes are integer pixel coordinates, x2 and y2 exclusive, up to 400 per request. Return {"data": [{"b64_json": "<base64 PNG>"}]} with exactly one 1280x720 8-bit RGB PNG (no alpha, no palette). Requests are sent one at a time.

Evaluation

Each image scores the mean of realism and quality, where quality is the geometric mean of OCR text accuracy, OCR layout F1 and faithfulness to the description. The score is the mean over 200 held-out images. An image scores zero, and the other images are scored normally, when its request still fails after the retries described under Submission, takes over 120 seconds (retries included) or returns an invalid image. The score is zero for the whole run if more than 1 request in 20 ends that way, p95 latency exceeds 30 seconds, or the 200 requests take over 7,200 seconds in total. All times are measured by the grading client inside the serving sandbox, so the grading system's own transport overhead never counts. Metric details are in /task/public/SCHEMA.md.

Artifacts

No starting model is provided. Record the checkpoint you started from in architecture.json as base_model and base_revision (an immutable commit), or "none" and "from-scratch" if you use no pretrained model. Keep your data download scripts and training logs with the submission: the review checks where your training data came from.

Include architecture.json with base_model, base_revision, training_evidence_files and inference_evidence_files. The file lists use paths relative to output/ and include training source, serve.sh and model-loading code. Keep initialization/training logs with the submission as readable text or JSON. Name the source of every training, calibration and selection corpus (in architecture.json or a data manifest) and keep its download commands and logs with the submission. I will check for training-data provenance. Ship compiled dependencies as unmodified PyPI or PyTorch wheels, installed into /tmp at startup. Prewarmed JIT kernel caches (Triton, TorchInductor) may be included.

Submission

Put serve.sh, code, weights and dependencies in /workspace/output/ as regular files. Relative symlinks that resolve inside output/ are copied as the regular files they point to. However, absolute, escaping, dangling or cyclic symlinks fail the submission (for example, use python -m venv --copies, or vendor wheels). The /workspace/output folder is copied to /submission when being evaluated. serve.sh runs as an unprivileged user that can write to /submission, /tmp and the preinstalled Python environment (/opt/venv) and nowhere else; install vendored wheels there or under /submission. You should package dependencies for offline installation, since the evaluation server will not have internet access. The command bash /submission/serve.sh receives HOST=0.0.0.0, PORT=8000 and MODEL_NAME=submission. GET /health must return 200 when ready, within 300 seconds. Startup and inference run offline on one H100 SXM 80GB, with 8 CPU cores and 80 GiB RAM. The serving image is Ubuntu 24.04 with the CUDA 12.8 runtime libraries (no CUDA compiler), Python 3.12 in /opt/venv first on PATH (pip, numpy, pillow, safetensors, httpx preinstalled), gcc and Python headers. Also preinstalled: torch==2.10.0+cu128, diffusers==0.40.0, transformers==5.17.0, accelerate==1.15.0, peft==0.21.0, safetensors==0.8.0, huggingface-hub==1.32.0, tokenizers==0.23.2. Vendor cp312 manylinux wheels for anything else. Every request must answer HTTP 200 with the documented JSON. A request that fails (non-200, connection error, invalid JSON) is retried twice on the same payload. If it still fails, or answers in an undocumented shape, you get no credit for that item. Up to one evaluation request in twenty may fail that way, but more than that gives you no credit for the entire run.

Development

The development machine is CPU-only. Use spin --help to rent compute and access model APIs, and spin budget to check how much allowance you have left. The API key your coding harness was started with is for the harness alone. Do not use the API key in your own code, scripts, curl or Spin machines, copy it anywhere, or start extra harness sessions as workers; AI model calls you make yourself must go through spin and spend your allowance. I will check your session transcript for this.

To check your solution locally, run python /task/validate.py --server-url URL. See /task/public for examples and supporting files.

You can check your solution on the same infrastructure I will evaluate it on using frontier selfcheck, which copies /workspace/output/, starts it as the evaluation does, sends public examples through the same response checks and measures the speed requirements on public inputs. Its report gives the stage reached, the digest of the files it checked, the GPU, the startup time and each requirement's measured value and margin, and the end of the server log. You can check your solution four times.

A check usually takes 10-40 minutes (frontier selfcheck status shows this task's limit) and runs while you keep working. See frontier selfcheck --help for more details.

When your solution is final, hand it in with frontier submit: it shows what /workspace/output/ holds and asks you to confirm with frontier submit --confirm, which submits the folder as it is and ends your session. Without a submission, /workspace/output/ is graded as it is when your session ends. See frontier submit --help.

Your session ends the moment you finish a turn without a tool call: background commands, scheduled wake-ups and reminders will not bring you back, and anything not yet in /workspace/output/ is lost. Wait for long jobs with foreground commands under the tool time limit (for example sleep 600; tail -n 3 train.log, repeated) and stop only when the final files are in place.