A Survey on Data Synthesis and Augmentation for Large Language Models
K Wang, J Zhu, M Ren, Z Liu, S Li, Z Zhang… - arXiv preprint arXiv …, 2024 - arxiv.org
The success of Large Language Models (LLMs) is inherently linked to the availability of vast,
diverse, and high-quality data for training and evaluation. However, the growth rate of high …
diverse, and high-quality data for training and evaluation. However, the growth rate of high …
Mmevol: Empowering multimodal large language models with evol-instruct
The development of Multimodal Large Language Models (MLLMs) has seen significant
advancements with increasing demands in various fields (eg, multimodal agents, embodied …
advancements with increasing demands in various fields (eg, multimodal agents, embodied …
[PDF][PDF] Baichuan-omni technical report
Y Li, H Sun, M Lin, T Li, G Dong, T Zhang… - arXiv preprint arXiv …, 2024 - researchgate.net
The salient multimodal capabilities and interactive experience of GPT-4o highlight its critical
role in practical applications, yet it lacks a high-performing open-source counterpart. In this …
role in practical applications, yet it lacks a high-performing open-source counterpart. In this …
Ocean-omni: To Understand the World with Omni-modality
Y Li, H Sun, M Lin, T Li, G Dong, T Zhang… - arXiv preprint arXiv …, 2024 - arxiv.org
The salient multimodal capabilities and interactive experience of GPT-4o highlight its critical
role in practical applications, yet it lacks a high-performing open-source counterpart. In this …
role in practical applications, yet it lacks a high-performing open-source counterpart. In this …
FM2DS: Few-Shot Multimodal Multihop Data Synthesis with Knowledge Distillation for Question Answering
Multimodal multihop question answering is a complex task that requires reasoning over
multiple sources of information, such as images and text, to answer questions. While there …
multiple sources of information, such as images and text, to answer questions. While there …
See it, Think it, Sorted: Large Multimodal Models are Few-shot Time Series Anomaly Analyzers
Time series anomaly detection (TSAD) is becoming increasingly vital due to the rapid growth
of time series data across various sectors. Anomalies in web service data, for example, can …
of time series data across various sectors. Anomalies in web service data, for example, can …
Rethinking Comprehensive Benchmark for Chart Understanding: A Perspective from Scientific Literature
Scientific Literature charts often contain complex visual elements, including multi-plot
figures, flowcharts, structural diagrams and etc. Evaluating multimodal models using these …
figures, flowcharts, structural diagrams and etc. Evaluating multimodal models using these …