FLANS at SemEval-2026 Task 7: RAG with Open-Sourced Smaller LLMs for Everyday Knowledge Across Diverse Languages and Cultures

2026-06-25 · Source: Paper Index on ACL Anthology · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics, Software Development & Engineering · Depth: Advanced, medium

Summary

The FLANS system participated in SemEval-2025 Task-7, "Everyday Knowledge Across Diverse Languages and Cultures," addressing both Short Answer Questions (SAQ) and Multiple-Choice Questions (MCQ) subtasks. This system employs Retrieval Augmented Generation (RAG) with open-sourced smaller LLMs (OS-sLLMs) deployed on the Ollama platform, prioritizing privacy and sustainability. A key component is its custom culturally aware knowledge base (CulKBs), built by extracting culturally-aware wiki-text and country-specific wiki-summaries from Wikipedia using prepared keyword lists. Additionally, one system variant integrates live online search output via DuckDuckGo. The team also developed and refined prompts, reporting their learning curve. The system was tested across English, Spanish, and Chinese languages, with all resources and code publicly shared.

Key takeaway

For Machine Learning Engineers developing multilingual RAG systems, consider integrating culturally aware knowledge bases and open-sourced smaller LLMs. You can enhance response accuracy by combining curated local data, like Wikipedia-derived CulKBs, with real-time online search via tools like DuckDuckGo. Deploying sLLMs on platforms such as Ollama offers a path to improved privacy and sustainability for your applications.

Key insights

RAG with OS-sLLMs and custom KBs enhances multilingual, culturally-aware everyday knowledge retrieval.

Principles

• Prioritize OS-sLLMs for privacy.
• Culturally-aware KBs improve RAG.
• Combine local KBs with live search.

Method

The system uses RAG with OS-sLLMs. It builds culturally aware knowledge bases (CulKBs) from Wikipedia via keyword extraction and integrates live online search. Prompts are refined and deployed on Ollama.

In practice

• Build custom KBs from Wikipedia.
• Integrate DuckDuckGo for live data.
• Deploy sLLMs on Ollama platform.

Topics

• Retrieval-Augmented Generation
• Open-Source Smaller LLMs (OS-sLLMs)
• Culturally Aware Knowledge Bases
• Multilingual NLP
• SemEval-2025 Task-7
• Ollama Platform
• Prompt Engineering

Code references

• aaronlifenghan/FLANS-2026

Best for: AI Engineer, Research Scientist, AI Scientist, Machine Learning Engineer, NLP Engineer

Related on AIssential

• Vector Search Using Ollama for Retrieval-Augmented Generation (RAG) — RAG combines vector search with LLMs to provide context-aware, fact-grounded responses from external data.
• SemEval-2026 Task 7: Everyday Knowledge Across Diverse Languages and Cultures — Evaluating LLM adaptability across diverse, low-resource language-culture pairs reveals critical performance and misalignment issues.
• Mod-Guide: An LLM-based Content Moderation Feedback System to Address Insensitive Speech toward Indigenous Ethnic and Religious Minority Communities — Integrating culturally grounded narratives via RAG enhances LLM sensitivity to minority perspectives in content moderation.
• Semantic adapters in text-to-SQL for low-resource languages: the importance of semantic information — Injecting semantic structural knowledge into LLMs significantly boosts Text-to-SQL performance for low-resource languages.
• Opportunities and Challenges of Large Language Models for Low-Resource Languages in Humanities Research — LLMs offer transformative potential for low-resource language research in humanities, despite significant data and ethical challenges.
• Teaching Language Models to Check Grounded Claim Factuality with Human Test-Taking Strategies — Formulating factuality checking as a true/false reading comprehension task with test-taking strategies improves LLM efficiency and accuracy.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Paper Index on ACL Anthology.