It is well-known that a diverse corpus is critical for training large language models, which are typically constructed from a mixture of various domains.In general, previous efforts resort to either sampling training data from different domains with static proportions or dynamically adjusting these proportions during training to optimise pretraining performance.However, few methods addressed the complexity of domain-adaptive continual pre-training.To fill this gap, we propose Velocitune, a novel framework that dynamically assesses learning velocity and adjusts data proportions accordingly, favouring slower learning domains while de-emphasising faster learning ones, which is guided by a scaling law to estimate the desired learning goal for each domain with a less associated cost.To evaluate the effectiveness of Velocitune, we conduct experiments on a dataset focused on reasoning tasks with CodeLlama, as well as on a corpus of system commands using Llama3 and Mistral.Velocitune achieves performance gains in both math and code reasoning tasks and command-line generation benchmarks.Further analysis reveals that key factors driving the effectiveness of Velocitune include target estimation and data ordering.* The first two authors contribute equally.

We propose ELAINE (EngLish-jApanese-chINesE)-medLLM, a trilingual (English, Japanese, Chinese) large language model adapted for the bio-medical domain based on Llama-3-8B. The training dataset was carefully curated in terms of volume and diversity to adapt to the biomedical domain and endow trilingual capability while preserving the knowledge and abilities of the base model. The training follows 2-stage paths: continued pre-training and supervised fine-tuning (SFT). Our results demonstrate that ELAINE-medLLM exhibits superior trilingual capabilities compared to existing bilingual or multilingual medical LLMs without severely sacrificing the base model’s capability.

LLMs have transformed NLP and shown promise in various fields, yet their potential in finance is underexplored due to a lack of comprehensive benchmarks, the rapid development of LLMs, and the complexity of financial tasks. In this paper, we introduce FinBen, the first extensive open-source evaluation benchmark, including 42 datasets spanning 24 financial tasks, covering eight critical aspects: information extraction (IE), textual analysis, question answering (QA), text generation, risk management, forecasting, decision-making, and bilingual (English and Spanish). FinBen offers several key innovations: a broader range of tasks and datasets, the first evaluation of stock trading, novel agent and Retrieval-Augmented Generation (RAG) evaluation, and two novel datasets for regulations and stock trading. Our evaluation of 21 representative LLMs, including GPT-4, ChatGPT, and the latest Gemini, reveals several key findings: While LLMs excel in IE and textual analysis, they struggle with advanced reasoning and complex tasks like text generation and forecasting. GPT-4 excels in IE and stock trading, while Gemini is better at text generation and forecasting. Instruction-tuned LLMs improve textual analysis but offer limited benefits for complex tasks such as QA. FinBen has been used to host the first financial LLMs shared task at the FinNLP-AgentScen workshop during IJCAI-2024, attracting 12 teams. Their novel solutions outperformed GPT-4, showcasing FinBen's potential to drive innovations in financial LLMs. All datasets and code are publicly available for the research community, with results shared and updated regularly on the Open Financial LLM Leaderboard.

Contrastive learning has recently been introduced into neural topic models (NTMs) to improve latent semantic discovery, but existing methods suffer from the sample bias problem owing to word frequency-based sampling strategy, which may result in false negative samples with similar semantics to the prototypes. We propose the novel graph contrastive neural topic model (GCTM), based on the graph-based sampling strategy, guided by the in-depth correlation and irrelevance information among documents and words. We model the input document as the document word bipartite graph (DWBG) and construct positive and negative word co-occurrence graphs (WCGs), to capture in-depth semantic correlation and irrelevance among words. Based on the DWBG and WCGs, we design the document-word information propagation (DWIP) process to perform the edge perturbation of DWBG, based on multi-hop correlations/irrelevance among documents and words. This yields the desired negative and positive samples, which are utilized for GCL together with the prototypes to improve learning document topic representations and latent topics. Experiments on several benchmark datasets demonstrate the effectiveness of our method for topic coherence and document representation learning compared with existing state-of-the-art methods.

Factual inconsistency with source documents in automatically generated summaries can lead to misinformation or pose risks. Existing factual consistency (FC) metrics are constrained by their performance, efficiency, and explainability. Recent advances in Large language models (LLMs) have demonstrated remarkable potential in text evaluation but their effectiveness in assessing FC in summarization remains underexplored. Prior research has mostly focused on proprietary LLMs, leaving essential factors that affect their assessment capabilities unexplored. Additionally, current FC evaluation benchmarks are restricted to news articles, casting doubt on the generality of the FC methods tested on them. In this paper, we first address the gap by introducing TreatFact—a dataset of LLM-generated summaries of clinical texts, annotated for FC by domain experts. Moreover, we benchmark 11 LLMs for FC evaluation across news and clinical domains and analyse the impact of model size, prompts, pre-training and fine-tuning data. Our findings reveal that despite proprietary models prevailing on the task, open-source LLMs lag behind. Nevertheless, there is potential for enhancing the performance of open-source LLMs through increasing model size, expanding pre-training data, and developing well-curated fine-tuning data. Experiments on TreatFact suggest that both previous methods and LLM-based evaluators are unable to capture factual inconsistencies in clinical summaries, posing a new challenge for FC evaluation.

Citation graphs can be helpful in generating high-quality summaries of scientific papers, where references of a scientific paper and their correlations can provide additional knowledge for contextualising its background and main contributions. Despite the promising contributions of citation graphs, it is still challenging to incorporate them into summarization tasks. This is due to the difficulty of accurately identifying and leveraging relevant content in references for a source paper, as well as capturing their correlations of different intensities. Existing methods either ignore references or utilize only abstracts indiscriminately from them, failing to tackle the challenge mentioned above. To fill that gap, we propose a novel citation-aware scientific paper summarization framework based on the citation graph, able to accurately locate and incorporate the salient contents from references, as well as capture varying relevance between source papers and their references. Specifically, we first build a domain-specific dataset PubMedCite with about 192K biomedical scientific papers and a large citation graph preserving 917K citation relationships between them. It is characterized by preserving the salient contents extracted from full texts of references, and the weighted correlation between the salient contents of references and the source paper. Based on it, we design a self-supervised citation-aware summarization framework (CitationSum) with graph contrastive learning, which boosts the summarization generation by efficiently fusing the salient information in references with source paper contents under the guidance of their correlations. Experimental results show that our model outperforms the state-of-the-art methods, due to efficiently leveraging the information of references and citation correlations.

Automatic disease named entity recognition (NER) plays a fundamental and essential role in knowledge extraction from biomedical literature. In this paper, we proposed a novel integrated model for disease mentions detection using deep network in combination with decoding algorithm and dictionary. To build the network, we implemented Bi-directional LSTM (Long Short-Term Memory) layers to capture long-term context information and fully-connected layers to improve the fitting capability, using concatenation of word embedding trained from raw biomedical texts and character embedding to encode the input. Viterbi algorithm was used to decode the previous output to access initial labeled sequence. On top of that, a disease names dictionary was constructed to label the disease mentions by exact string matching, which provided extra information to optimize the initial output. While training and testing on NCBI disease corpus, our model achieved F-score of 89.58% which performed better than current reported systems.