Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Aster: Autonomous Scientific Discovery over 20x Faster Than Existing Methods

Emmett Bicker

TL;DR

An AI agent for autonomous scientific discovery capable of operating over 20 times faster than existing frameworks, Aster's significant reduction in the number of iterations required for novel discovery expands the domain of tractable problems to include tasks with long evaluation durations, such as multi-hour machine learning training runs.

Abstract

We introduce Aster, an AI agent for autonomous scientific discovery capable of operating over 20 times faster than existing frameworks. Given a task, an initial program, and a script to evaluate the performance of the program, Aster iteratively improves the program, often leading to new state-of-the-art performances. Aster's significant reduction in the number of iterations required for novel discovery expands the domain of tractable problems to include tasks with long evaluation durations, such as multi-hour machine learning training runs. We applied Aster to problems in mathematics, GPU kernel engineering, biology, neuroscience, and language model training. More specifically: the Erdos minimum overlap problem, optimizing the TriMul kernel, a single-cell analysis denoising problem, training a neural activity prediction model to perform well on ZAPBench, and the NanoGPT Speedrun Competition. Aster attains SOTA results in every task, except for ZAPBench, where it matches the performance of the best human solution with less than 1/190th of the compute. Aster is accessible via a web interface and API at asterlab.ai.

Aster: Autonomous Scientific Discovery over 20x Faster Than Existing Methods

TL;DR

An AI agent for autonomous scientific discovery capable of operating over 20 times faster than existing frameworks, Aster's significant reduction in the number of iterations required for novel discovery expands the domain of tractable problems to include tasks with long evaluation durations, such as multi-hour machine learning training runs.

Abstract

We introduce Aster, an AI agent for autonomous scientific discovery capable of operating over 20 times faster than existing frameworks. Given a task, an initial program, and a script to evaluate the performance of the program, Aster iteratively improves the program, often leading to new state-of-the-art performances. Aster's significant reduction in the number of iterations required for novel discovery expands the domain of tractable problems to include tasks with long evaluation durations, such as multi-hour machine learning training runs. We applied Aster to problems in mathematics, GPU kernel engineering, biology, neuroscience, and language model training. More specifically: the Erdos minimum overlap problem, optimizing the TriMul kernel, a single-cell analysis denoising problem, training a neural activity prediction model to perform well on ZAPBench, and the NanoGPT Speedrun Competition. Aster attains SOTA results in every task, except for ZAPBench, where it matches the performance of the best human solution with less than 1/190th of the compute. Aster is accessible via a web interface and API at asterlab.ai.
Paper Structure (1 section, 8 figures, 4 tables)

This paper contains 1 section, 8 figures, 4 tables.

Table of Contents

  1. Programs

Figures (8)

  • Figure 1: Discoveries found by Aster across Mathematics, Kernel Engineering, Biology, and ML. All baseline results cited from TTT-Discoveryuksekgonul2026learningdiscovertesttime unless cited otherwise.
  • Figure 2: System Overview: Aster iteratively refines programs, evaluates their performance, stores them in a database, and repeats.
  • Figure 3: Visual comparison of circle packing solutions.
  • Figure 4: Comparison of Erdős Minimum Overlap constructions. Right: Aster's construction. Left: TTT-Discover.
  • Figure 5: Trajectory of the best program found for Single-Cell Denoising.
  • ...and 3 more figures