Constarium
← Search

Data · dataset · 2024

Open Source Conversational LLMs knowledge about Spanish words (100 words)

Listed in e-cienciaDatos

Dataset with a sample of 100 Spanish words and their definition and two examples of usage given by LLM models.

Description

Frequencies_100_words_CREA.xlsx: usage frequency of the words in Spanish llm_*.xlsx: responses from the LLM models about if they know the word, their definition, and two examples of sentences using them. It includes the manual validation from an expert in Spanish language.

LLMs tested: Llama-2-7b-chat-hf (quantization 32 bit) Llama-2-13b-chat-hf (quantization 16 bits) Llama-2-70b-chat-hf (quantization 4 bits) Mistral-7b-Instruct (quantization 32 bits) Mixtral-8x7b-Instruct (quantization 4 bits) Gemma-7b-it (quantization 32 bits) SOLAR-10.7b-Instruct (quantization 16 bits) Yi-6b-Chat (quantization 32 bits) Yi-34b-Chat (quantization 8 bits) Bloomz-7b1 (quantization 32 bits) FLOR-6.3b-Instructed (quantization 32 bits) Bertin-6b (quantization 32 bits) validation_*.xlsx: authomatic validation made by gpt3.5 and gpt4 of the llm_responses_100_words

Links

Where it is published

Catalogue records · 1
Provenance · 1 source records, 5 field assertions
SourceKeyLast seenRaw
e-cienciaDatosdoi:10.5281/ZENODO.107980959 d agoJSON v1
FieldAssertionExtractorEvidence
created_datesource · edatos consorciomadrono esconnector:edatos_consorciomadrono_es@1.0.0
descriptionsource · edatos consorciomadrono esconnector:edatos_consorciomadrono_es@1.0.0/description
publication_datesource · edatos consorciomadrono esconnector:edatos_consorciomadrono_es@1.0.0
titlesource · edatos consorciomadrono esconnector:edatos_consorciomadrono_es@1.0.0/name
updated_datesource · edatos consorciomadrono esconnector:edatos_consorciomadrono_es@1.0.0