Research map / Large language models / Retrieval and knowledge
Measuring Meta-Cultural Competency: A Spectral Framework for LLM Knowledge Structures
Sougata Saha, Madhur Jindal, Saurabh Kumar Pandey, Mahardika Krisna Ihsani, Alham Fikri Aji, Monojit Choudhury
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use an instruction-tuned model instead of an older model (GPT-2, GPT-J-6B) | macrostructure evaluation · Macro-F1 vs human regimes | macro-F1 | Improved |
| scale Llama from 8B to 70B | Llama-3.1-it · Macro-F1 vs human regimes | macro-F1 | No clear effect |
| scale Gemma-2 from 2B to 9B | Gemma-2-it · Macro-F1 vs human regimes | macro-F1 | No clear effect |
| add explicit nudge to ask clarifying questions (ablation) | Llama-3.1-8B/70B-it recommenders · recipe recommendation simulation | APR, INT | No clear effect |
| add explicit nudge to ask clarifying questions (ablation) | Aya-8B recommender · recipe recommendation simulation | APR, INT | Improved |
| add explicit nudge to ask clarifying questions (ablation) | Llama-3.2-1B-it recommender · recipe recommendation simulation | APR, INT | No clear effect |
| add explicit nudge to ask clarifying questions (ablation) | Gemma-2 recommenders · recipe recommendation simulation | APR | No clear effect |
| rank models by macrostructural alignment rather than microstructural benchmarks | model ranking · predicting downstream APR | Spearman correlation: APR: 0.942; INT: 0.828 | Improved |
| swap judge from GPT-4o to Gemini-3-Flash-Preview (ablation) | LLM-as-judge · 100 sampled conversations | correlation: 0.84 for INT and 0.76 for APR | No clear effect |
| compare microstructural benchmark rankings with macrostructural rankings | model ranking · cultural evaluation | ranking | Mixed |
Closest papers by meaning
- CulturePark: Boosting Cross-cultural Understanding in Large Language Models (NeurIPS 2024 · similarity 0.90)
- CultureLLM: Incorporating Cultural Differences into Large Language Models (NeurIPS 2024 · similarity 0.88)
- KoLA: Carefully Benchmarking World Knowledge of Large Language Models (ICLR 2024 · similarity 0.88)
- BertaQA: How Much Do Language Models Know About Local Culture? (NeurIPS 2024 · similarity 0.87)
- BLEnD: A Benchmark for LLMs on Everyday Knowledge in Diverse Cultures and Languages (NeurIPS 2024 · similarity 0.87)
- Culture In a Frame: C$^3$B as a Comic-Based Benchmark for Multimodal Culturally Awareness (ICLR 2026 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
