Search

Your search keyword '"Osmulski P"' showing total 6 results
Start Over Author "Osmulski P" Publication Year Range This year
6 results on '"Osmulski P"'

1. Enhancing Q&A Text Retrieval with Ranking Models: Benchmarking, fine-tuning and deploying Rerankers for RAG

Author

Moreira, Gabriel de Souza P., Ak, Ronay, Schifferer, Benedikt, Xu, Mengyao, Osmulski, Radek, and Oldridge, Even

Subjects
Computer Science - Information Retrieval, Computer Science - Computation and Language, Computer Science - Machine Learning
Abstract

Ranking models play a crucial role in enhancing overall accuracy of text retrieval systems. These multi-stage systems typically utilize either dense embedding models or sparse lexical indices to retrieve relevant passages based on a given query, followed by ranking models that refine the ordering of the candidate passages by its relevance to the query. This paper benchmarks various publicly available ranking models and examines their impact on ranking accuracy. We focus on text retrieval for question-answering tasks, a common use case for Retrieval-Augmented Generation systems. Our evaluation benchmarks include models some of which are commercially viable for industrial applications. We introduce a state-of-the-art ranking model, NV-RerankQA-Mistral-4B-v3, which achieves a significant accuracy increase of ~14% compared to pipelines with other rerankers. We also provide an ablation study comparing the fine-tuning of ranking models with different sizes, losses and self-attention mechanisms. Finally, we discuss challenges of text retrieval pipelines with ranking models in real-world industry applications, in particular the trade-offs among model size, ranking accuracy and system requirements like indexing and serving latency / throughput., Comment: Accepted for the 1st Workshop on GenAI and RAG Systems for Enterprise @ CIKM 2024

Published
2024

2. NV-Retriever: Improving text embedding models with effective hard-negative mining

Author

Moreira, Gabriel de Souza P., Osmulski, Radek, Xu, Mengyao, Ak, Ronay, Schifferer, Benedikt, and Oldridge, Even

Subjects
Computer Science - Information Retrieval, Computer Science - Artificial Intelligence
Abstract

Text embedding models have been popular for information retrieval applications such as semantic search and Question-Answering systems based on Retrieval-Augmented Generation (RAG). Those models are typically Transformer models that are fine-tuned with contrastive learning objectives. Many papers introduced new embedding model architectures and training approaches, however, one of the key ingredients, the process of mining negative passages, remains poorly explored or described. One of the challenging aspects of fine-tuning embedding models is the selection of high quality hard-negative passages for contrastive learning. In this paper we propose a family of positive-aware mining methods that leverage the positive relevance score for more effective false negatives removal. We also provide a comprehensive ablation study on hard-negative mining methods over their configurations, exploring different teacher and base models. We demonstrate the efficacy of our proposed methods by introducing the NV-Retriever-v1 model, which scores 60.9 on MTEB Retrieval (BEIR) benchmark and 0.65 points higher than previous methods. The model placed 1st when it was published to MTEB Retrieval on July 07, 2024.

Published
2024

4. PAI-1 uncouples integrin-β1 from restrain by membrane-bound β-catenin to promote collagen fibril remodeling in obesity-related neoplasms

Author

Lin, Li-Ling, Nayak, Bijaya, Osmulski, Pawel A., Wang, Exing, Wang, Chen-Pin, Valente, Philip T., Wang, Chiou-Miin, Tan, Xi, Santanam, Nalini, Wang, Tian-Li, Gaczynska, Maria E., Kost, Edward R., Huang, Tim H.-M., and Kirma, Nameer B.

Abstract

The paracrine actions of adipokine plasminogen activator inhibitor-1 (PAI-1) are implicated in obesity-associated tumorigenesis. Here, we show that PAI-1 mediates extracellular matrix (ECM) signaling via epigenetic repression of DKK1in endometrial epithelial cells (EECs). While the loss of DKK1 is known to increase β-catenin accumulation for WNT signaling activation, this epigenetic repression causes β-catenin release from transmembrane integrins. Furthermore, PAI-1 elicits the disengagement of TIMP2 and SPARC from integrin-β1 on the cell surface, lifting an integrin-β1-ECM signaling constraint. The heightened interaction of integrin-β1 with type 1 collagen (COL1) remodels extracellular fibrillar structures in the ECM. Consequently, the enhanced nanomechanical stiffness of this microenvironment is conducive to EEC motility and neoplastic transformation. The formation of extensively branched COL1 fibrils is also observed in endometrial tumors of patients with obesity. The findings highlight PAI-1 as a contributor to enhanced integrin-COL1 engagement and extensive ECM remodeling during obesity-associated neoplastic development.

Published
2024
Full Text
View/download PDF

5. β-Amyloid impairs Proteasome structure and function. Proteasome activation mitigates amyloid induced toxicity and cognitive deficits.

Author

Davidson K, Bano M, Parker D, Osmulski P, Gaczynska M, and Pickering AM

Abstract

Background: Alzheimer's Disease (AD) is the leading cause of dementia globally, affecting around 50 million people and marked by cognitive decline and the accumulation of β-amyloid plaques and hyperphosphorylated tau. The limited treatment options and numerous failed clinical trials targeting β-amyloid (Aβ) highlight the need for novel approaches. Lowered proteasome activity is a consistent feature in AD, particularly in the hippocampus. Impaired proteasome function in AD is hypothesized to stem from direct inhibition by β-amyloid or hyperphosphorylated tau, disrupting critical neuronal processes such as memory formation and synaptic plasticity., Objectives: This study tests the hypothesis that AD related deficits are driven in part by impaired proteasome function as a consequence of inhibition by Aβ. We evaluated how proteasome function is modulated by Aβ and the capacity of two proteasome-activating compounds, TAT1-8,9-TOD and TAT1-DEN to rescue Aβ-induced impairment in vitro, as well as survival deficits in cell culture and Aβ-induced cognitive deficits in Drosophila and mouse models., Results: Our study demonstrates that oligomeric β-amyloid binds to the 20S proteasome and impairs its activity and conformational stability. The oligomers also destabilize the 26S proteasome to release the free 20S proteasome. Treatment with proteasome activators TAT1-8,9TOD and TAT1-DEN rescue the 20S proteasome function and reduces cell death caused by Aβ42 toxicity in SK-N-SH cells. In Drosophila models overexpressing Aβ42, oral administration of proteasome agonists delayed mortality and restored cognitive function. Chronic treatment with TAT1-DEN protected against deficits in working memory caused by Aβ42 in mice and in hAPP(J20) mice with established deficits, acute TAT1-DEN treatment significantly improved spatial learning, with treated mice performing comparably to controls., Conclusions: Aβ has dual impacts on 20S and 26S proteasome function and stability. Proteasome activation using TAT1-8,9TOD and TAT1-DEN shows promise in mitigating AD-like deficits by protecting against amyloid toxicity and enhancing proteasome function. These findings suggest that targeting proteasome activity could be a viable therapeutic approach for AD, warranting further investigation into the broader impacts of proteasome modulation on AD pathology.

Published
2024
Full Text
View/download PDF

6. Targeting S6K/NFκB/SQSTM1/Polθ signaling to suppress radiation resistance in prostate cancer.

Author

Clark A, Villarreal MR, Huang SB, Jayamohan S, Rivas P, Hussain SS, Ybarra M, Osmulski P, Gaczynska ME, Shim EY, Smith T, Gupta YK, Yang X, Delma CR, Natarajan M, Lai Z, Wang LJ, Michalek JE, Higginson DS, Ikeno Y, Ha CS, Chen Y, Ghosh R, and Kumar AP

Subjects
Male, Animals, Humans, Mice, Cell Line, Tumor, DNA-Directed DNA Polymerase genetics, DNA-Directed DNA Polymerase metabolism, Ribosomal Protein S6 Kinases, 70-kDa metabolism, Ribosomal Protein S6 Kinases, 70-kDa genetics, Prostatic Neoplasms radiotherapy, Prostatic Neoplasms pathology, Prostatic Neoplasms metabolism, Prostatic Neoplasms genetics, Prostatic Neoplasms drug therapy, Sequestosome-1 Protein metabolism, Sequestosome-1 Protein genetics, NF-kappa B metabolism, NF-kappa B genetics, Signal Transduction drug effects, Radiation Tolerance drug effects
Abstract

In this study we have identified POLθ-S6K-p62 as a novel druggable regulator of radiation response in prostate cancer. Despite significant advances in delivery, radiotherapy continues to negatively affect treatment outcomes and quality of life due to resistance and late toxic effects to the surrounding normal tissues such as bladder and rectum. It is essential to develop new and effective strategies to achieve better control of tumor. We found that ribosomal protein S6K (RPS6KB1) is elevated in human prostate tumors, and contributes to resistance to radiation. As a downstream effector of mTOR signaling, S6K is known to be involved in growth regulation. However, the impact of S6K signaling on radiation response has not been fully explored. Here we show that loss of S6K led to formation of smaller tumors with less metastatic ability in mice. Mechanistically we found that S6K depletion reduced NFκB and SQSTM1 (p62) reporter activity and DNA polymerase θ (POLθ) that is involved in alternate end-joining repair. We further show that the natural compound berberine interacts with S6K in a in a hitherto unreported novel mode and that pharmacological inhibition of S6K with berberine reduces Polθ and downregulates p62 transcriptional activity via NFκB. Loss of S6K or pre-treatment with berberine improved response to radiation in prostate cancer cells and prevented radiation-mediated resurgence of PSA in animals implanted with prostate cancer cells. Notably, silencing POLQ in S6K overexpressing cells enhanced response to radiation suggesting S6K sensitizes prostate cancer cells to radiation via POLQ. Additionally, inhibition of autophagy with CQ potentiated growth inhibition induced by berberine plus radiation. These observations suggest that pharmacological inhibition of S6K with berberine not only downregulates NFκB/p62 signaling to disrupt autophagic flux but also decreases Polθ. Therefore, combination treatment with radiation and berberine inhibits autophagy and alternate end-joining DNA repair, two processes associated with radioresistance leading to increased radiation sensitivity., Competing Interests: Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper., (Copyright © 2024 The Author(s). Published by Elsevier B.V. All rights reserved.)

Published
2024
Full Text
View/download PDF

Catalog

Books, media, physical & digital resources
See catalog results