Skip to main content
PRISM measures a model’s ability to learn from scratch. Miners submit two scripts, a model architecture.py and a custom training.py loop. The challenge owns the dataset and the evaluation: the validator re-executes the miner’s training loop under a forced random initialization with a fixed seed, and computes the score itself, ignoring anything the miner reports. This page is a high-level overview. For quickstart, submission format, scoring detail, constraints, examples, and the API, see the dedicated PRISM tab.

PRISM tab

Quickstart, submission format, scoring, constraints, and the PRISM API.

How a submission flows

1

Submit two scripts

A miner submits a bundle: architecture.py exposes build_model(ctx) and training.py exposes train(ctx).
2

Validate and sandbox

PRISM validates the two-script contract and runs the static AST sandbox.
3

LLM hard gate

An OpenRouter LLM reviews both scripts as a hard gate and can reject before any GPU work.
4

Forced-init re-execution

The validator re-executes the training loop on a GPU under a forced random init on the locked FineWeb-Edu train split.
5

Score and weight

The challenge computes the prequential bits-per-byte score plus a held-out delta tie-breaker, then converts ranks into normalized dry-run weights.

Scoring basis

The score is a prequential (online) compression metric in bits-per-byte (bpb): the area under the from-scratch loss curve, normalized by the raw bytes of text consumed. A model that learns faster compresses the stream better and earns a better score.

Anti-cheat by construction

  • No pretrained weights: forced random init makes smuggled weights inert, and the container runs with no network.
  • No metric manipulation: the challenge re-executes and computes the metric itself; miner-reported numbers and manifests are ignored.
  • No memorization: the validation and test splits are secret; an excessive train-versus-held-out gap penalizes the score.

Agent Challenge

The other primary challenge on BASE.

All challenges

Every challenge running on the subnet.

Source

The challenge lives in its own repository: BASE/prism.