Skip to content

Anarres

Measure and compare training environments before committing the full training budget.

Anarres is an open-source toolkit and research program for determining which training environments are most likely to improve a model at fixed compute. It connects domain-native measurements to portable evidence without flattening the science that produced them.

Install Anarres, then follow the first steps or browse the released research programs.

Highlights

  • Covers mathematical reasoning, cyber, robotics, legal reasoning, molecular design, materials science, and agentic finance through native formats.
  • Separates task qualification, starting-policy measurement, and validation against later training outcomes, while keeping infrastructure failures distinct from ordinary failures.
  • Preserves measurement definitions, units, uncertainty, costs, and provenance in versioned, content-addressed records.
  • Combines a fast Rust CLI for shared operations with installable adapters for specialized domain runtimes.

Installation

Install the native CLI, then verify that it is available:

$ brew install anarres-organization/tap/anarres
$ anarres --version

$ curl --proto '=https' --tlsv1.2 -LsSf \
    https://github.com/anarres-organization/anarres/releases/latest/download/anarres-cli-installer.sh | sh
$ anarres --version

powershell -ExecutionPolicy ByPass -c "irm https://github.com/anarres-organization/anarres/releases/latest/download/anarres-cli-installer.ps1 | iex"
anarres --version

Then follow the first steps, or see every installation method in the release guide. Domain adapters are installed only when a workflow needs its specialized runtime.

Quick start

Run the bundled CPU-only workflow:

$ anarres demo

  ✓ Qualified    math-reference-demo-v1
  ✓ Measured     math-reference-demo-v1
  ✓ Compatible   candidate panel

ENVIRONMENT               SIGNAL / 1M   COVERAGE  ADMITTED   REPAIR
math-alternative-demo-v1     214.2857     0.4154         2        1
math-reference-demo-v1       833.3333     0.6533         2        1

The demo qualifies two environments, measures both under the same synthetic starting policy, checks that their records are comparable, and writes validated records to .anarres/demo/.

To use native data, connect a source to its matching adapter and inspect the requirements before running a policy probe:

$ anarres init ./candidate-a --adapter <adapter> \
    --revision <immutable-revision> --license <SPDX-identifier>
$ anarres inspect ./candidate-a
$ anarres qualify ./candidate-a

The first-run workflow continues from a native source to a portable environment record. The domain support matrix shows which formats, probes, and outcome tests are available today.

How it works

native task sources
    → qualify tasks, verifiers, and rewards
    → measure experience under the starting policy
    → compare compatible candidate environments
    → test frozen predictions through controlled training

Anarres keeps three questions separate: whether a task and verifier work, whether the starting policy produces usable experience, and whether training on that experience improves a held-out target. Infrastructure failures never become ordinary failures.

The native CLI handles routing, validation, hashing, and comparison. Installable adapters retain specialized runtimes where simulators, evaluators, chemistry packages, or executable sandboxes are required.

Domain workflows

Program Available workflow Evidence boundary
Mathematics Native grouped-probe processor Four-scale environment-utility evidence
Formal verification Vero runtime qualification Execution study continues separately
Robotics Can source qualification and candidate measurement Prospective PushT and Can evidence
Legal reasoning LAB qualification and legal regression reports SylloGym and LAB evidence remain separate
Materials science OMat24 qualification and MLIP measurements Scored prospective MLIP test
Molecular design Reaction-bank qualification and dry-run probe Matched training outcomes remain unopened
Vulnerability exploitation Sanitized CyberGym qualification No environment-comparison result
Agentic finance Business Bench qualification Retrospective evidence; no prospective outcome

Learn more