What it measures

Classic pressure test inserting target facts ("needles") at varying document depths (0–100%) and context lengths (4k–2M tokens).

Reading the measurement

Length is a test condition, not a capability score. Locate the relevant evidence, then examine whether the model resolves conflicts and combines details correctly. Keep document extraction, input length and answer grading fixed.

Published evaluation

Text NIAH · 128k; HELMET paper · 128k text retrieval. The source’s task split, model configuration and scoring units are retained.

Where this fits in an evaluation suite

A large context window is a capacity specification. Successful retrieval and cross-document reasoning are observed capabilities. Test both at the actual lengths your workflow needs.

Start with a capability hypothesis, then choose a second test with a different failure mode. Keep the checkpoint, reasoning effort, agent tools, sampling policy, and scoring procedure fixed when measuring a training intervention. Record failures and missing outputs in the denominator.

A reproducible comparison

Text NIAH · 128k; HELMET paper · 128k text retrieval. The source’s task split, model configuration and scoring units are retained.

Publish the exact run configuration, item coverage, score units, and available uncertainty. If a source supplies only an aggregate or a single run, preserve that limit. A new model name or a rounded score alone does not establish a capability improvement.

PRIMARY EVIDENCE & FURTHER READING

  1. HELMET: How to Evaluate Long-Context Language Models Effectively and Thoroughly — published result table ↗

    Howard Yen, Tianyu Gao, Minmin Hou, Ke Ding, Daniel Fleischer, Peter Izsak et al. · 2024

    All 8 authors

    Howard Yen, Tianyu Gao, Minmin Hou, Ke Ding, Daniel Fleischer, Peter Izsak, Moshe Wasserblat, Danqi Chen

  2. Needle In A Multimodal Haystack — published result table ↗

    Weiyun Wang, Shuibo Zhang, Yiming Ren, Yuchen Duan, Tiantong Li, Shuo Liu et al. · 2024

    All 16 authors

    Weiyun Wang, Shuibo Zhang, Yiming Ren, Yuchen Duan, Tiantong Li, Shuo Liu, Mengkang Hu, Zhe Chen, Kaipeng Zhang, Lewei Lu, Xizhou Zhu, Ping Luo, Yu Qiao, Jifeng Dai, Wenqi Shao, Wenhai Wang

Reviewed 2026-10-26. Explanation and interpretation by RLTasks. Benchmark ownership remains with its original authors.