9 papers
InfoLaw: Information Scaling Laws for Large Language Models with Quality-Weighted Mixture Data and Repetition
Fengze Liu, Weidong Zhou, Binbin Liu +7
Upweighting high-quality data in LLM pretraining often improves performance, but in datalimited regimes, especially under overtraining, stronger upweighting increases repetition an…
Target-Oriented Pretraining Data Selection via Neuron-Activated Graph
Zijun Wang, Haoqin Tu, Weidong Zhou +7
Everyday tasks come with a target, and pretraining models around this target is what turns them into experts. In this paper, we study target-oriented language model (LM) pretrainin…
Exploring Polyglot Harmony: On Multilingual Data Allocation for Large Language Models Pretraining
Ping Guo, Yubing Ren, Binbin Liu +6
Large language models (LLMs) have become integral to a wide range of applications worldwide, driving an unprecedented global demand for effective multilingual capabilities. Central…
TiKMiX: Take Data Influence into Dynamic Mixture for Language Model Pre-training
Yifan Wang, Binbin Liu, Fengze Liu +6
The data mixture used in the pre-training of a language model is a cornerstone of its final performance. However, a static mixing strategy is suboptimal, as the model's learning pr…
MuRating: A High Quality Data Selecting Approach to Multilingual Large Language Model Pretraining
Zhixun Chen, Ping Guo, Wenhan Han +10
Data quality is a critical driver of large language model performance, yet existing model-based selection methods focus almost exclusively on English. We introduce MuRating, a scal…
MuBench: Assessment of Multilingual Capabilities of Large Language Models Across 61 Languages
Wenhan Han, Yifan Zhang, Zhixun Chen +7
Multilingual large language models (LLMs) are advancing rapidly, with new models frequently claiming support for an increasing number of languages. However, existing evaluation dat…
MoORE: SVD-based Model MoE-ization for Conflict- and Oblivion-Resistant Multi-Task Adaptation
Shen Yuan, Yin Zheng, Taifeng Wang +2
Adapting large-scale foundation models in multi-task scenarios often suffers from task conflict and oblivion. To mitigate such issues, we propose a novel ''model MoE-ization'' stra…
QuaDMix: Quality-Diversity Balanced Data Selection for Efficient LLM Pretraining
Fengze Liu, Weidong Zhou, Binbin Liu +8
Quality and diversity are two critical metrics for the training data of large language models (LLMs), positively impacting performance. Existing studies often optimize these metric…
LogicMP: A Neuro-symbolic Approach for Encoding First-order Logic Constraints
Weidi Xu, Jingwei Wang, Lele Xie +7
Integrating first-order logic constraints (FOLCs) with neural networks is a crucial but challenging problem since it involves modeling intricate correlations to satisfy the constra…