reasoning-gym

Procedural reasoning generators with deterministic algorithmic scorers; only reviewed families are served.

Field	Value
Reference	nodus/reasoning-gym@1.0.0
Image	nodus/env-reasoning-gym:1.0.0
Publisher	Open-Thought
Category	Reasoning
Readiness	Research
Modes	Train, Evaluate
Reward	Scalar
Held-out measures	TrainedTask
Splits	2000 train, 320 test (disjoint by canonical identity)
Licenses	code Apache-2.0, data Apache-2.0
Source	https://github.com/open-thought/reasoning-gym

Grading

Completions are graded by the platform, never by the trainer: the trainer submits {taskId, completion} batches and the verdicts come back as task events.

Grader	Kind	What it checks
score-answer	Program	The family’s score_answer; only a full score is Correct, the score is the reward

Sample task

One line of nodus-env tasks --split test --seed 0; tasks never carry answers.

{
  "metadata": {
    "family": "basic_arithmetic"
  },
  "prompt": "Calculate -3 - 8 / ( -8 + 1 + 8 ).\n\nReply with the final answer only. Do not restate the question, do not show working, and do not add any words, labels or punctuation around it.",
  "taskId": "reasoning-gym:test:0:0"
}

Examples

Each example is a TrainingJob template. A baseline is shown only where the example was measured by running it.

Example	Mode	Runtime	Model	Tasks	Baseline	Trained	Measured on
chain-sum	Evaluate	nodus/evaluate	Qwen/Qwen3-1.7B @ 70d244c	64	not measured	n/a (evaluation)	not measured
number-format	Evaluate	nodus/evaluate	Qwen/Qwen3-1.7B @ 70d244c	64	not measured	n/a (evaluation)	not measured
basic-arithmetic	Evaluate	nodus/evaluate	Qwen/Qwen3-1.7B @ 70d244c	64	not measured	n/a (evaluation)	not measured
products	Evaluate	nodus/evaluate	Qwen/Qwen3-1.7B @ 70d244c	64	not measured	n/a (evaluation)	not measured
letter-counting	Evaluate	nodus/evaluate	Qwen/Qwen3-1.7B @ 70d244c	64	not measured	n/a (evaluation)	not measured

Run one with a server dry-run first:

$ nodus create trainingjob my-run --from-example nodus/reasoning-gym:chain-sum --dry-run=server -o estimate

import nodus

job = nodus.recipes.TrainingJob.from_example("nodus/reasoning-gym:chain-sum")
plan = job.preview()
run = plan.run(max_cost=5)
print(run.wait().summary)