LingBench++: A Linguistically-Informed Benchmark and Reasoning Framework for Multi-Step and Cross-Cultural Inference with LLMs

TL;DR

LingBench++ introduces a comprehensive benchmark and reasoning framework for evaluating LLMs on complex, cross-cultural linguistic tasks with structured reasoning and rich metadata, aiming to improve interpretability and cultural awareness.

Contribution

It presents LingBench++, a novel benchmark with structured reasoning traces and metadata, and a multi-agent architecture integrating grammatical knowledge and iterative reasoning for improved LLM performance.

Findings

Models with external knowledge and iterative reasoning outperform single-pass models.

LingBench++ enables evaluation of LLMs on low-resource and cross-cultural languages.

Structured reasoning improves interpretability and accuracy.

Abstract

We propose LingBench++, a linguistically-informed benchmark and reasoning framework designed to evaluate large language models (LLMs) on complex linguistic tasks inspired by the International Linguistics Olympiad (IOL). Unlike prior benchmarks that focus solely on final answer accuracy, LingBench++ provides structured reasoning traces, stepwise evaluation protocols, and rich typological metadata across over 90 low-resource and cross-cultural languages. We further develop a multi-agent architecture integrating grammatical knowledge retrieval, tool-augmented reasoning, and deliberate hypothesis testing. Through systematic comparisons of baseline and our proposed agentic models, we demonstrate that models equipped with external knowledge sources and iterative reasoning outperform single-pass approaches in both accuracy and interpretability. LingBench++ offers a comprehensive foundation for advancing linguistically grounded, culturally informed, and cognitively plausible reasoning in LLMs.