2024/10/19 by Mingxin Li, Zhijie Nie, Li, Mingxin +9 · 2 citations
Computer Science · #Computation and Language (cs.CL) #FOS: Computer and information sciences #Intelligent Tutoring Systems and Adaptive Learning #Topic Modeling
paper · pdf · doi:10.48550/arxiv.2410.15035
openalex publication_date 2024/10/19 · openalex created_date 2024/11/06 · openalex updated_date 2026/07/28
Text embeddings are vital for tasks such as text retrieval and semantic textual similarity (STS). Recently, the advent of pretrained language models, along with unified benchmarks like the Massive Text Embedding Benchmark (MTEB), has facilitated the development of versatile general-purpose text embedding models. Advanced embedding models are typically developed using large-scale multi-task data and joint training across multiple tasks. However, our experimental analysis reveals two significant drawbacks of joint training: 1) Task Conflict: Gradients from different tasks interfere with each other, leading to negative transfer. 2) Data Imbalance: Disproportionate data distribution introduces biases that negatively impact performance across tasks. To overcome these challenges, we explore model merging-a technique that combines independently trained models to mitigate gradient conflicts and balance data distribution. We introduce a novel method, Self Positioning, which efficiently searches for optimal model combinations within the interpolation space of task vectors using stochastic gradient descent. Our experiments demonstrate that Self Positioning significantly enhances multi-task performance on the MTEB dataset, achieving an absolute improvement of 0.7 points. It outperforms traditional resampling methods while reducing computational costs. This work offers a robust approach to building generalized text embedding models with superior performance across diverse embedding-related tasks.