Full-Parameter Continual Pretraining of Gemma2: Insights into Fluency and Domain Knowledge

• URL: http://arxiv.org/abs/2505.05946v2
• Date: Thu, 05 Jun 2025 11:13:42 GMT
• Title: Full-Parameter Continual Pretraining of Gemma2: Insights into Fluency and Domain Knowledge
• Authors: Vytenis Šliogeris, Povilas Daniušis, Artūras Nakvosas,
• Abstract summary: We empirically investigate the relationship between linguistic fluency and domain knowledge in the context of continual learning with large language models (LLMs)<n>Specifically, we enhance the linguistic fluency of the Gemma2 LLM for the Lithuanian language by autoregressively pretraining its full parameter set on the first 10% of the Lithuanian language component of the CulturaX dataset.<n>To prevent catastrophic forgetting of the model's existing domain knowledge, we apply Elastic Weight Consolidation (EWC)<n>In the post-training evaluations, we assess linguistic fluency through perplexity and evaluate domain knowledge using accuracy on a suite of language understanding benchmarks.
• Score: 0.0
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: In this technical report, we empirically investigate the relationship between linguistic fluency and domain knowledge in the context of continual learning with large language models (LLMs). Specifically, we enhance the linguistic fluency of the Gemma2 LLM for the Lithuanian language by autoregressively pretraining its full parameter set on the first 10\% of the Lithuanian language component of the CulturaX dataset. To prevent catastrophic forgetting of the model's existing domain knowledge, we apply Elastic Weight Consolidation (EWC), leveraging Fisher information estimated using data from the Massive Multitask Language Understanding (MMLU) benchmark. In the post-training evaluations, we assess linguistic fluency through perplexity and evaluate domain knowledge using accuracy on a suite of language understanding benchmarks, including ARC-Easy, Belebele, GSM8K, HellaSwag, MMLU, TruthfulQA, and Winogrande, in both English and Lithuanian. The empirical results demonstrate that EWC not only mitigates catastrophic forgetting by preserving the model's performance in terms of both linguistic fluency and domain knowledge but also improves or maintains these capabilities for the newly added Lithuanian language. These findings highlight the potential for more efficient adaptation of general-purpose LLMs to under-represented languages without requiring access to the original training data. The accompanying codebase is openly accessible at https://github.com/Neurotechnology/LLM_EWC.

Related papers

• Aligning Knowledge Graphs and Language Models for Factual Accuracy [7.205708660952737]
  We introduce ALIGNed-LLM, a simple yet effective approach to improve language models' factuality.<n>We use embeddings from a pre-trained Knowledge Graph Embedding (KGE) model, such as TransE, and a trainable projection layer to align entity and text embeddings.
  arXiv  Detail & Related papers  (2025-07-17T08:15:50Z)
• Evaluating Large Language Model with Knowledge Oriented Language Specific Simple Question Answering [73.73820209993515]
  We introduce KoLasSimpleQA, the first benchmark evaluating the multilingual factual ability of Large Language Models (LLMs)<n>Inspired by existing research, we created the question set with features such as single knowledge point coverage, absolute objectivity, unique answers, and temporal stability.<n>Results show significant performance differences between the two domains.
  arXiv  Detail & Related papers  (2025-05-22T12:27:02Z)
• Open or Closed LLM for Lesser-Resourced Languages? Lessons from Greek [2.3499129784547663]
  We evaluate the performance of open-source (Llama-70b) and closed-source (GPT-4o mini) large language models on seven core NLP tasks with dataset availability.<n>Second, we expand the scope of Greek NLP by reframing Authorship Attribution as a tool to assess potential data usage by LLMs in pre-training.<n>Third, we showcase a legal NLP case study, where a Summarize, Translate, and Embed (STE) methodology outperforms the traditional TF-IDF approach for clustering emphlong legal texts.
  arXiv  Detail & Related papers  (2025-01-22T12:06:16Z)
• Assessing Language Comprehension in Large Language Models Using Construction Grammar [3.0906699069248806]
  Construction Grammar (CxG) provides insights into the meaning captured by linguistic elements known as constructions (Cxns)<n>These datasets are carefully constructed to include examples which are unlikely to appear in pre-training data, yet intuitive and easy for humans to understand.<n>Our experiments focus on downstream natural language inference and reasoning tasks by comparing LLMs' understanding of the underlying meanings communicated through 8 unique Cxns with that of humans.
  arXiv  Detail & Related papers  (2025-01-08T18:15:10Z)
• Adaptive Pruning for Large Language Models with Structural Importance Awareness [66.2690963378878]
  Large language models (LLMs) have significantly improved language understanding and generation capabilities.<n>LLMs are difficult to deploy on resource-constrained edge devices due to their high computational and storage resource demands.<n>We propose structurally-aware adaptive pruning (SAAP) to significantly reduce the computational and memory costs while maintaining model performance.
  arXiv  Detail & Related papers  (2024-12-19T18:08:04Z)
• Language Portability Strategies for Open-domain Dialogue with Pre-trained Language Models from High to Low Resource Languages [1.7436854281619139]
  We propose a study of linguistic portability strategies of large pre-trained language models (PLMs) used for open-domain dialogue systems. In particular the target low-resource language (L_T) will be simulated with French, as it lacks of task-specific resources.
  arXiv  Detail & Related papers  (2024-07-01T14:20:54Z)
• Crosslingual Capabilities and Knowledge Barriers in Multilingual Large Language Models [62.91524967852552]
  Large language models (LLMs) are typically multilingual due to pretraining on diverse multilingual corpora.<n>But can these models relate corresponding concepts across languages, i.e., be crosslingual?<n>This study evaluates state-of-the-art LLMs on inherently crosslingual tasks.
  arXiv  Detail & Related papers  (2024-06-23T15:15:17Z)
• Bridging the Bosphorus: Advancing Turkish Large Language Models through Strategies for Low-Resource Language Adaptation and Benchmarking [1.3716808114696444]
  Large Language Models (LLMs) are becoming crucial across various fields, emphasizing the urgency for high-quality models in underrepresented languages. This study explores the unique challenges faced by low-resource languages, such as data scarcity, model selection, evaluation, and computational limitations.
  arXiv  Detail & Related papers  (2024-05-07T21:58:45Z)
• The Power of Question Translation Training in Multilingual Reasoning: Broadened Scope and Deepened Insights [108.40766216456413]
  We propose a question alignment framework to bridge the gap between large language models' English and non-English performance. Experiment results show it can boost multilingual performance across diverse reasoning scenarios, model families, and sizes. We analyze representation space, generated response and data scales, and reveal how question translation training strengthens language alignment within LLMs.
  arXiv  Detail & Related papers  (2024-05-02T14:49:50Z)
• FAC$^2$E: Better Understanding Large Language Model Capabilities by Dissociating Language and Cognition [56.76951887823882]
  Large language models (LLMs) are primarily evaluated by overall performance on various text understanding and generation tasks. We present FAC$2$E, a framework for Fine-grAined and Cognition-grounded LLMs' Capability Evaluation.
  arXiv  Detail & Related papers  (2024-02-29T21:05:37Z)
• Embracing Language Inclusivity and Diversity in CLIP through Continual Language Learning [58.92843729869586]
  Vision-language pre-trained models (VL-PTMs) have advanced multimodal research in recent years, but their mastery in a few languages like English restricts their applicability in broader communities. We propose to extend VL-PTMs' language capacity by continual language learning (CLL), where a model needs to update its linguistic knowledge incrementally without suffering from catastrophic forgetting (CF) We construct a CLL benchmark covering 36 languages based on MSCOCO and XM3600 datasets and then evaluate multilingual image-text retrieval performance.
  arXiv  Detail & Related papers  (2024-01-30T17:14:05Z)
• KGLens: Towards Efficient and Effective Knowledge Probing of Large Language Models with Knowledge Graphs [22.53643028991214]
  Large Language Models (LLMs) might hallucinate facts, while curated Knowledge Graph (KGs) are typically factually reliable. Measuring the alignment between KGs and LLMs can effectively probe the factualness and identify the knowledge blind spots of LLMs. We present KGLens, a Thompson-inspired framework aimed at effectively and efficiently measuring the alignment between KGs and LLMs.
  arXiv  Detail & Related papers  (2023-12-15T23:34:05Z)
• Tokenizer Choice For LLM Training: Negligible or Crucial? [30.33170936148845]
  We study the influence of tokenizer choice on Large Language Models (LLMs) downstream performance by training 24 mono- and multilingual LLMs. We find that the tokenizer choice can significantly impact the model's downstream performance and training costs. We show that multilingual tokenizers trained on the five most frequent European languages require vocabulary size increases of factor three in comparison to English.
  arXiv  Detail & Related papers  (2023-10-12T22:44:19Z)
• Extrapolating Large Language Models to Non-English by Aligning Languages [109.09051737966178]
  Existing large language models show disparate capability across different languages. In this paper, we empower pre-trained LLMs on non-English languages by building semantic alignment across languages.
  arXiv  Detail & Related papers  (2023-08-09T13:32:06Z)
• Improving Domain-Specific Retrieval by NLI Fine-Tuning [64.79760042717822]
  This article investigates the fine-tuning potential of natural language inference (NLI) data to improve information retrieval and ranking. We employ both monolingual and multilingual sentence encoders fine-tuned by a supervised method utilizing contrastive loss and NLI data. Our results point to the fact that NLI fine-tuning increases the performance of the models in both tasks and both languages, with the potential to improve mono- and multilingual models.
  arXiv  Detail & Related papers  (2023-08-06T12:40:58Z)
• Mixed-Lingual Pre-training for Cross-lingual Summarization [54.4823498438831]
  Cross-lingual Summarization aims at producing a summary in the target language for an article in the source language. We propose a solution based on mixed-lingual pre-training that leverages both cross-lingual tasks like translation and monolingual tasks like masked language models. Our model achieves an improvement of 2.82 (English to Chinese) and 1.15 (Chinese to English) ROUGE-1 scores over state-of-the-art results.
  arXiv  Detail & Related papers  (2020-10-18T00:21:53Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.