Testuen analisia

Benchmarking Critical Questions Generation: A Challenging Reasoning Task for Large Language Models

The task of Critical Questions Generation (CQs-Gen) aims to foster critical thinking by enabling systems to generate questions that expose underlying assumptions and challenge the validity of argumentative reasoning structures. Despite growing interest in this area, progress has been hindered by the lack of suitable datasets and automatic evaluation standards. This paper presents a comprehensive approach to support the development and benchmarking of systems for this task. We construct the first large-scale dataset including ~5K manually annotated questions.

CRITICS - Critical Science Without Borders: LLMs for Translation of Scientific Knowledge in Multilingual Contexts

"Muga Gabeko Zientzia Kritikoa" CRITICS proiektuak ezagutza zientifikorako sarbidea demokratizatzea du helburu, eduki zientifikoaren itzulpen automatikorako Hizkuntza-Eredu Handi espezializatuak garatuz tokiko hizkuntzetan. Ekimenak hiru berrikuntza nagusi konbinatzen ditu: dokumentu zientifikoetarako IA sistema aurreratuak, tokiko testuinguruetara edukia egokitzen duten IAk sortutako Irakaskuntza-Ikaskuntza Sekuentziak, eta ikasleen ama hizkuntzetan pentsamendu kritikoa ebaluatzeko ebaluazio-tresna automatizatuak.

DeepThought: Reasoning and Alignment of Large Language Models for Multilingual and Low Resource Scenarios

DeepThought-ek eskala handiko hurbilketa bat aurkezten du hizkuntza gutxituetarako Hizkuntza Eredu Handiak egokitzeko, prestakuntza eta lerrokadura estrategia berritzaile baten bidez. Proiektuak enpirikoki balioztatutako metodologiak garatuko ditu Llama-3+ eta Qwen+ bezalako kode irekiko ereduak euskara eta gaztelaniara zabaltzeko, lehendik dauden LLMen eta Hizkuntza Arrazonamendu Ereduen datu-multzo sintetikoak eta arrazonamendu-berbalizazioak aprobetxatuz.

Humanizing AI with language technology (HumanAIze)

The "Humanizing AI with Language Technology (HumanAIze)" project seeks to enhance Large Language Models (LLMs) to make them more human, reliable, and ethical by addressing weaknesses such as bias, cultural understanding, and reliability. Six research centers bring their expertise in areas such as multilingualism, anonymization, verification, and evaluation. A key approach is "reliable by design," implementing ethical and legal principles, data governance, human feedback, and sustainability.

EFICIENCIA DE MODELOS LLM PARA INDUSTRIAS ESTRATÉGICAS (EMIE) (2025.0736)

HAZITEK ESTRATEGIKOA. EMIE. RVCTI AZPIKONTRATAZIOA LLMak industrian duten erabilera ebaluatzeko proiektua. RVCTI azpikontratazioa HAZITEK estrategiko batean.

Truth Knows No Language: Evaluating Truthfulness Beyond English

We introduce a professionally translated extension of the TruthfulQA benchmark designed to evaluate truthfulness in Basque, Catalan, Galician, and Spanish. Truthfulness evaluations of large language models (LLMs) have primarily been focused on English. However, the ability of LLMs to maintain truthfulness across languages remains under-explored. Our study evaluates 12 state-of-the-art open LLMs, comparing base and instruction-tuned models using human evaluation, multiple-choice metrics, and LLM-as-a-Judge scoring.

Orriak

RSS - Testuen analisia-rako harpidetza egin