NoLiMa

reasoning official site →

Multilingual: no

NoLiMa (No Literal Matching) is a long-context benchmark extending needle-in-a-haystack tests with minimal lexical overlap between questions and needles, requiring models to infer latent associations rather than relying on surface-level matching. Published at ICML 2025.

Methodology

Imported from llm-stats public benchmark metadata. Modality: text. Max score: 1. Categories: long_context, reasoning. Language: en. Verified by llm-stats: no.