Skip to content

NGSS-aligned middle school and high school science benchmarks

Product or Programme · All products and programmes
Evidence · no claims cited

Description

Two publicly released synthetic benchmark datasets created by Schroeder et al. for evaluating LLM performance on NGSS-aligned middle- and high-school science items.

Components

  • NGSS-aligned middle school and high school science benchmarks with a four-tier synthetic data quality gate pipeline: Two publicly released ground-truth benchmarks for evaluating LLM understanding of secondary science: "Our final dataset consisted of 1,078 items for middle school benchmark, and 1,150 items for the high school benchmark." Items were synthetically generated by Sonnet 4.6 Thinking, distributed across recall, conceptual understanding, and applied reasoning types and across 12 NGSS areas per level. Construction used "a four-tier data quality gate system" removing duplicates, enforcing Flesch-Kincaid reading-level ranges, and requiring a three-judge LLM panel to agree on both answers and standard alignment. Data and code are available on GitHub. (Schroeder et al. (2026))

Claims

Key Sources

  • Schroeder, N. L., Ambarwati, Y. E., Zhang, Y., & Zhai, C. (2026). A Benchmark for LLM's Understanding of Middle School and High School Science. https://github.com/InviteInstitute/MS-NGSS-Aligned-Benchmark
Page metadata
Field Value
type product
id ngss-aligned-middle-school-and-high-school-science-benchmarks
title NGSS-aligned middle school and high school science benchmarks
description Two publicly released synthetic benchmark datasets created by Schroeder et al.
status review
product_kind dataset
generated by process:settle-candidates on 2026-10-11
trust unverified — no one has confirmed this page's content

Frontmatter is the page's machine-readable record — see the Schema & Guide.