arXiv subjects

Liana Mikaelyan

Publications and source records attributed to Liana Mikaelyan.

4 recordsLinked to original sources

arXiv2025

DeltaLLM: Compress LLMs with Low-Rank Deltas between Shared Weights

We introduce DeltaLLM, a new post-training compression technique to reduce the memory footprint of LLMs. We propose an alternative way of structuring LLMs with weight sharing between layers in subsequent Transformer blocks, along with additional low-rank difference matrices between them. For training, we adopt the progressing module replacement method and show that the lightweight training of the low-rank modules with approximately 30M-40M tokens is sufficient to achieve performance on par with LLMs of comparable sizes trained from scratch. We release the resultant models, DeltaLLAMA and DeltaPHI, with a 12% parameter reduction, retaining 90% of the performance of the base Llama and Phi models on common knowledge and reasoning benchmarks. Our method also outperforms compression techniques JointDrop, LaCo, ShortGPT and SliceGPT with the same number of parameters removed. For example, DeltaPhi 2.9B with a 24% reduction achieves similar average zero-shot accuracies as recovery fine-tuned SlicedPhi 3.3B with a 12% reduction, despite being approximately 400M parameters smaller with no fine-tuning applied. This work provides new insights into LLM architecture design and compression methods when storage space is critical.

cs.LG↗

arXiv2025

KBLaM: Knowledge Base augmented Language Model

In this paper, we propose Knowledge Base augmented Language Model (KBLaM), a new method for augmenting Large Language Models (LLMs) with external knowledge. KBLaM works with a knowledge base (KB) constructed from a corpus of documents, transforming each piece of knowledge in the KB into continuous key-value vector pairs via pre-trained sentence encoders with linear adapters and integrating them into pre-trained LLMs via a specialized rectangular attention mechanism. Unlike Retrieval-Augmented Generation, KBLaM eliminates external retrieval modules, and unlike in-context learning, its computational overhead scales linearly with KB size rather than quadratically. Our approach enables integrating a large KB of more than 10K triples into an 8B pre-trained LLM of only 8K context window on one single A100 80GB GPU and allows for dynamic updates without model fine-tuning or retraining. Experiments demonstrate KBLaM's effectiveness in various tasks, including question-answering and open-ended reasoning, while providing interpretable insights into its use of the augmented knowledge. Code and datasets are available at https://github.com/microsoft/KBLaM/

cs.AI↗

arXiv2024

Team MTS @ AutoMin 2021: An Overview of Existing Summarization Approaches and Comparison to Unsupervised Summarization Techniques

Remote communication through video or audio conferences has become more popular than ever because of the worldwide pandemic. These events, therefore, have provoked the development of systems for automatic minuting of spoken language leading to AutoMin 2021 challenge. The following paper illustrates the results of the research that team MTS has carried out while participating in the Automatic Minutes challenge. In particular, in this paper we analyze existing approaches to text and speech summarization, propose an unsupervised summarization technique based on clustering and provide a pipeline that includes an adapted automatic speech recognition block able to run on real-life recordings. The proposed unsupervised technique outperforms pre-trained summarization models on the automatic minuting task with Rouge 1, Rouge 2 and Rouge L values of 0.21, 0.02 and 0.2 on the dev set, with Rouge 1, Rouge 2, Rouge L, Adequacy, Grammatical correctness and Fluency values of 0.180, 0.035, 0.098, 1.857, 2.304, 1.911 on the test set accordingly

cs.CL↗

arXiv2024

Learning to Extract Structured Entities Using Language Models

Recent advances in machine learning have significantly impacted the field of information extraction, with Language Models (LMs) playing a pivotal role in extracting structured information from unstructured text. Prior works typically represent information extraction as triplet-centric and use classical metrics such as precision and recall for evaluation. We reformulate the task to be entity-centric, enabling the use of diverse metrics that can provide more insights from various perspectives. We contribute to the field by introducing Structured Entity Extraction and proposing the Approximate Entity Set OverlaP (AESOP) metric, designed to appropriately assess model performance. Later, we introduce a new Multistage Structured Entity Extraction (MuSEE) model that harnesses the power of LMs for enhanced effectiveness and efficiency by decomposing the extraction task into multiple stages. Quantitative and human side-by-side evaluations confirm that our model outperforms baselines, offering promising directions for future advancements in structured entity extraction. Our source code and datasets are available at https://github.com/microsoft/Structured-Entity-Extraction.

cs.CL↗