Saved in:
Bibliographic Details
Main Author: Yang, Chengshuai
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2603.25780
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866918411432361984
author Yang, Chengshuai
author_facet Yang, Chengshuai
contents Large language models can generate scientific simulation code, but the generated code silently fails on most non-textbook problems. We show that classical mathematical validation -- well-posedness, convergence, and error certification -- can be fully automated by a Judge Agent, reducing the silent-failure rate from 42% to 1.5% across 134 test cases spanning 12 scientific domains. The headline result comes from a prospective benchmark: 72 blinded tasks submitted by 12 independent scientists yield an 89% success rate (95% CI: [80%, 95%]) with automated error bounds, versus 53% without the Judge. On clinical CT (the only powered experiment, n = 200), the pipeline reaches 99% of expert quality. The residual 1.5% concentrates at bifurcation points where certifiability breaks down. We formalize this boundary through the simulability class S and introduce spec.md, a structured specification format that makes any scientific computation problem machine-readable and solver-independent. Code, data, and all 72 benchmark tasks are publicly archived.
format Preprint
id arxiv_https___arxiv_org_abs_2603_25780
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle A Judge Agent Closes the Reliability Gap in AI-Generated Scientific Simulation
Yang, Chengshuai
Software Engineering
Machine Learning
68T07, 65N15, 65M15, 65Y15
Large language models can generate scientific simulation code, but the generated code silently fails on most non-textbook problems. We show that classical mathematical validation -- well-posedness, convergence, and error certification -- can be fully automated by a Judge Agent, reducing the silent-failure rate from 42% to 1.5% across 134 test cases spanning 12 scientific domains. The headline result comes from a prospective benchmark: 72 blinded tasks submitted by 12 independent scientists yield an 89% success rate (95% CI: [80%, 95%]) with automated error bounds, versus 53% without the Judge. On clinical CT (the only powered experiment, n = 200), the pipeline reaches 99% of expert quality. The residual 1.5% concentrates at bifurcation points where certifiability breaks down. We formalize this boundary through the simulability class S and introduce spec.md, a structured specification format that makes any scientific computation problem machine-readable and solver-independent. Code, data, and all 72 benchmark tasks are publicly archived.
title A Judge Agent Closes the Reliability Gap in AI-Generated Scientific Simulation
topic Software Engineering
Machine Learning
68T07, 65N15, 65M15, 65Y15
url https://arxiv.org/abs/2603.25780