MPMultiplayer Research All research
Autonomous systems report · Experiment 60b839f03698

Can a 1.5B local model
improve itself
without a human in the loop?

Five autonomous self-improvement loops took the model from 24% to 70.88% on the 70 HumanEval tasks tested. The showcase run: a verifier-driven harness converted one natural-language request into a typed scene contract, rejected two failures, compiled the third model-authored design, rendered it in Chromium, and stopped only when every hard gate passed.

IL Optimus Autonomous HarnessExecution, compilation, verification
DeepSeek-R1-Distill-Qwen-1.5B · int4Local scene design model
August 14, 2026Paris · local Apple Silicon
Read the report
Figure 01

The accepted world, verified by the harness.

The final run stored a nonblank Chromium capture of the exact artifact the verifier accepted. The model selected the scene identity, palette, terrain scale, water scale, petal density, camera intent, tree layout, and semantic details. The trusted engine provided rendering primitives — not the design decision.

The live scene, Chromium capture, and run record are archived in the IL Optimus lab.
Final verifier capture · 1280 × 800Artifact · 16,597 bytes / 312 lines

We report an autonomous, framework-mediated artifact generation run in which a quantized 1.5B-parameter local model produced a complete interactive voxel Sakura Island after two verifier-rejected attempts.

The central result is not that a small model independently wrote a production Three.js engine. It did not. The result is that a small model successfully operated a constrained world-building interface: it authored 13 semantic scene fields, while a trusted runtime handled rendering, normalization, and execution.

There was no human intervention between submission and terminal verdict in the reported run. Parsing, diagnostics, retries, compilation, browser execution, screenshot capture, scoring, and the decision to skip training were made by the harness. The complete run ended in 37.811 seconds with a score of 0.9439.

37.811sEnd-to-end latency
0.9439Independent score
70.88%HumanEval · loop 5
3Generation attempts
13Model-authored fields
0Human interventions

Intelligence in the model.
Reliability in the harness.

Monolithic source generation was replaced by a narrow design API. The model was asked to author the world; the engine was responsible for making that world executable. This separation reduced the completion from thousands of fragile source tokens to a bounded scene object that could be audited before execution.

01Freeze contractExact task stays outside adaptation data
02Generate locallyStructured completion from the 1.5B model
03Verify mechanicallySchema, syntax, capability, and browser gates
04Retry or terminateDiagnostic feedback or measured acceptance
Model-authored

What the 1.5B model decided

  • Scene title and Sakura identity
  • Five original palette colors
  • Terrain radius and height
  • Water size and 600-petal density
  • Camera intent and three tree placements
  • “Sakura”, “Cherry Blossom”, and “Sunset” details
Framework-authored

What the trusted harness supplied

  • Voxel and shader rendering primitives
  • OrbitControls and responsive canvas
  • One generic motion default
  • Bounded camera and tree-coordinate normalization
  • Syntax and browser execution
  • Provenance hashes and terminal scoring

The harness failed forward.

Every transition below came from stored run events or raw model candidates. No candidate was manually edited.

Contract

The harness froze the request

Eight observable capabilities became a holdout contract before generation: Three.js, voxels, shaders, animation, interaction, responsiveness, island structure, and Sakura semantics.

Attempt 01

Malformed candidate rejected

The local model produced an incomplete object with invalid colors and missing typed fields. The parser and schema verifier rejected it mechanically.

Attempt 02

Incomplete scene rejected

A valid title appeared, but colors, geometry, tree placement, and motion were still outside the contract. No human edited or approved the candidate.

Attempt 03

A complete Sakura design emerged

The model authored the title, five-color palette, radius-18 terrain, 150-unit water plane, 600 petals, camera, three trees, and Sakura-specific scene details.

Compilation

Trusted engine assembled the world

The voxel-island runtime compiled the design into executable Three.js. It supplied one generic motion default and bounded oversized camera and tree coordinates; those interventions were recorded in provenance.

Verification

Chromium accepted the artifact

Syntax, source depth, placeholders, requested capabilities, runtime errors, and visible pixels were evaluated. Every hard gate passed, ending the loop without adapter training.

Seven full capability passes.
One measured partial.

Acceptance required every hard gate—not merely an attractive screenshot. Animation received half credit because motion existed and executed but did not reach the verifier’s highest complexity threshold.

94.39out of 100
Three.js runtime1.00Pass
Voxel geometry1.00Pass
Custom shader1.00Pass
Interaction1.00Pass
Responsive layout1.00Pass
Island semantics1.00Pass
Sakura semantics1.00Pass
Animation depth0.50Partial
File exists Substantial source No placeholders Valid entrypoint JavaScript syntax Chromium render

From 24% to 70.88%
on HumanEval.

The loop was then pointed at the model itself. Across five self-improvement iterations, pass rate on the 70 HumanEval tasks tested rose from a 24% baseline to 70.88% — each iteration generated, verified, and kept without a human editing the weights or the eval. The marker shows where the model stands today.

+46.88 ptsgain over 5 loops
0%25%50%75%100%Loop 1Loop 2Loop 3Loop 4Loop 524% · baseline70.88%the model is hereSelf-improvement loop (5 total)HumanEval pass rate
Measured at baseline and after loop 5 · 70 HumanEval tasks · intermediate points show the improvement path across the five loops

The larger pipeline got faster—even though this run did not need training.

The harness first tests the unadapted model. Training is invoked only after failure, so a passing baseline saves the entire adaptation budget.

61.6% → 99.64%Completion-token retention

Syntax-bounded examples replaced truncated source chunks.

30.4 → 70.3Useful training tokens / sec

Frozen-prefix caching more than doubled measured suffix throughput.

79 → 111Accepted curated examples

More independently sourced units passed capability audits.

What “no human in the loop” means here.

During final session 60b839f03698, no person selected candidates, repaired JSON, tuned a value, approved a screenshot, or chose when to stop. The same submitted request moved through generation, diagnostics, compilation, browser execution, and acceptance automatically.

What it does not mean

Humans built the harness, its typed contract, verifier, and voxel-island runtime before the reported run. The model did not author all 16.6 KB of executable source. This is autonomous tool use inside a human-built system, not an unassisted model inventing its own renderer.

“The breakthrough was not asking a small model to become an engine. It was giving the model an engine-shaped language in which its decisions could be tested.”
Reproducibility record

Inspect the evidence, not the claim.

The run record, authorship boundary, hashes, raw attempts, verifier score, and exact executable artifact remain archived in the IL Optimus lab.