evolvent-ai / RSIGym

RSIGym: A Flexible Environment for Recursive Self-Improvement 
[![Evolvent AI][evolvent-image]][evolvent-url]
[![Paper][paper-image]][paper-url]
[![Website][website-image]][website-url]
[![Records][records-image]][records-url]
[![GitHub][github-image]][github-url]
[![License][license-image]](LICENSE)
[![X][x-image]][x-url]
[![小红书][xhs-image]][xhs-url]
An **agent-native · service-based · budget-controlled** environment for AI research. RSIGym gives a research agent reusable tools to generate data, train models, evaluate candidates, and revise execution harnesses. The agent chooses the experiments; the platform handles the infrastructure, permissions, and service budget.
**RSI-Index** measures how much of the remaining performance gap those experiments close across five benchmark domains.
## ✨ Features
- **Everything as a Service** — Training, inference, rollout, evaluation, and sandbox execution are available through reusable interfaces. Agents work in lightweight, CPU-only containers and call the services when needed.
- **Three improvement tracks** — Study training data, execution harnesses, or their joint optimization through configurable permissions.
- **A shared research budget** — Each run receives one platform key with a service budget and access rules. Authorization checks and cost accounting are shared across services.
- **Evidence for the next experiment** — Training metrics, per-task rewards, execution logs, and trajectories help agents diagnose failures and choose what to try next.
- **Inspectable research runs** — Task definitions specify the initial system, allowed changes, submission artifacts, evaluation protocol, and post-run audits.
## 🏗 How it works
 [code] 
The **research agent** designs and runs experiments. The **target system** is the model and execution harness it improves. These can be different models, or the same model in a self-improvement experiment.
### Services
| Component | What it provides | Implementation |
|---|---|---|
| Training | LoRA fine-tuning, training metrics, checkpoints | train_server, backed by Tinker |
| Inference | OpenAI-compatible access to base models and trained checkpoints | model_server, backed by Tinker |
| Rollout | Frontier-model calls for data generation and judging | rollout_server, through an OpenAI-compatible upstream |
| Evaluation | Benchmark jobs, task rewards, trajectories, and logs | benchmark_server, using Harbor and E2B |
| Sandbox | Isolated computation through the E2B SDK | e2b_proxy |
| Authorization | Run keys, permissions, and service-budget accounting | auth_server |
### Improvement tracks
| Track | The agent can change | Held fixed |
|---|---|---|
| **Data** | Training data | Training settings and harness |
| **Harness** | Execution harness | Model weights |
| **Joint** | Training data, training settings, and harness | Starting base model and evaluation protocol |
The same service infrastructure supports all three tracks. Each task's `key.json` defines its budget and service