LIVIVO - The Search Portal for Life Sciences

zur deutschen Oberfläche wechseln
Advanced search

Search results

Result 1 - 4 of total 4

Search options

  1. Book ; Online: Poisoning Language Models During Instruction Tuning

    Wan, Alexander / Wallace, Eric / Shen, Sheng / Klein, Dan

    2023  

    Abstract: Instruction-tuned LMs such as ChatGPT, FLAN, and InstructGPT are finetuned on datasets that contain user-submitted examples, e.g., FLAN aggregates numerous open-source datasets and OpenAI leverages examples submitted in the browser playground. In this ... ...

    Abstract Instruction-tuned LMs such as ChatGPT, FLAN, and InstructGPT are finetuned on datasets that contain user-submitted examples, e.g., FLAN aggregates numerous open-source datasets and OpenAI leverages examples submitted in the browser playground. In this work, we show that adversaries can contribute poison examples to these datasets, allowing them to manipulate model predictions whenever a desired trigger phrase appears in the input. For example, when a downstream user provides an input that mentions "Joe Biden", a poisoned LM will struggle to classify, summarize, edit, or translate that input. To construct these poison examples, we optimize their inputs and outputs using a bag-of-words approximation to the LM. We evaluate our method on open-source instruction-tuned LMs. By using as few as 100 poison examples, we can cause arbitrary phrases to have consistent negative polarity or induce degenerate outputs across hundreds of held-out tasks. Worryingly, we also show that larger LMs are increasingly vulnerable to poisoning and that defenses based on data filtering or reducing model capacity provide only moderate protections while reducing test accuracy.

    Comment: ICML 2023
    Keywords Computer Science - Computation and Language ; Computer Science - Cryptography and Security ; Computer Science - Machine Learning
    Subject code 006
    Publishing date 2023-05-01
    Publishing country us
    Document type Book ; Online
    Database BASE - Bielefeld Academic Search Engine (life sciences selection)

    More links

    Kategorien

  2. Article ; Online: Retained Catheter in the Aorta.

    Huang, Alex / Quesada, Daniel / Aguìñiga-Navarrete, Phillip / Rosbrugh, James / Wan, Alexander

    Clinical practice and cases in emergency medicine

    2019  Volume 3, Issue 4, Page(s) 434–435

    Abstract: Due to the recent increase in endovascular procedures, retained foreign bodies such as stents and catheters in vasculature have become a common and serious complication. Treatments for these complications vary depending on the acuity and stability of the ...

    Abstract Due to the recent increase in endovascular procedures, retained foreign bodies such as stents and catheters in vasculature have become a common and serious complication. Treatments for these complications vary depending on the acuity and stability of the foreign body in the vessel. We discuss a rare case of an adult found to have an incidental retained umbilical artery catheter in the aorta.
    Language English
    Publishing date 2019-09-30
    Publishing country United States
    Document type Journal Article
    ISSN 2474-252X
    ISSN (online) 2474-252X
    DOI 10.5811/cpcem.2019.7.43425
    Database MEDical Literature Analysis and Retrieval System OnLINE

    More links

    Kategorien

  3. Book ; Online: GLUECons

    Faghihi, Hossein Rajaby / Nafar, Aliakbar / Zheng, Chen / Mirzaee, Roshanak / Zhang, Yue / Uszok, Andrzej / Wan, Alexander / Premsri, Tanawan / Roth, Dan / Kordjamshidi, Parisa

    A Generic Benchmark for Learning Under Constraints

    2023  

    Abstract: Recent research has shown that integrating domain knowledge into deep learning architectures is effective -- it helps reduce the amount of required data, improves the accuracy of the models' decisions, and improves the interpretability of models. However, ...

    Abstract Recent research has shown that integrating domain knowledge into deep learning architectures is effective -- it helps reduce the amount of required data, improves the accuracy of the models' decisions, and improves the interpretability of models. However, the research community is missing a convened benchmark for systematically evaluating knowledge integration methods. In this work, we create a benchmark that is a collection of nine tasks in the domains of natural language processing and computer vision. In all cases, we model external knowledge as constraints, specify the sources of the constraints for each task, and implement various models that use these constraints. We report the results of these models using a new set of extended evaluation criteria in addition to the task performances for a more in-depth analysis. This effort provides a framework for a more comprehensive and systematic comparison of constraint integration techniques and for identifying related research challenges. It will facilitate further research for alleviating some problems of state-of-the-art neural models.

    Comment: 8 pages, Accepted in AAAI 2023 proceedings
    Keywords Computer Science - Machine Learning ; Computer Science - Artificial Intelligence ; Computer Science - Computation and Language
    Subject code 004
    Publishing date 2023-02-16
    Publishing country us
    Document type Book ; Online
    Database BASE - Bielefeld Academic Search Engine (life sciences selection)

    More links

    Kategorien

  4. Article ; Online: Deficiency of terminal ADP-ribose protein glycohydrolase TARG1/C6orf130 in neurodegenerative disease.

    Sharifi, Reza / Morra, Rosa / Appel, C Denise / Tallis, Michael / Chioza, Barry / Jankevicius, Gytis / Simpson, Michael A / Matic, Ivan / Ozkan, Ege / Golia, Barbara / Schellenberg, Matthew J / Weston, Ria / Williams, Jason G / Rossi, Marianna N / Galehdari, Hamid / Krahn, Juno / Wan, Alexander / Trembath, Richard C / Crosby, Andrew H /
    Ahel, Dragana / Hay, Ron / Ladurner, Andreas G / Timinszky, Gyula / Williams, R Scott / Ahel, Ivan

    The EMBO journal

    2013  Volume 32, Issue 9, Page(s) 1225–1237

    Abstract: Adenosine diphosphate (ADP)-ribosylation is a post-translational protein modification implicated in the regulation of a range of cellular processes. A family of proteins that catalyse ADP-ribosylation reactions are the poly(ADP-ribose) (PAR) polymerases ( ...

    Abstract Adenosine diphosphate (ADP)-ribosylation is a post-translational protein modification implicated in the regulation of a range of cellular processes. A family of proteins that catalyse ADP-ribosylation reactions are the poly(ADP-ribose) (PAR) polymerases (PARPs). PARPs covalently attach an ADP-ribose nucleotide to target proteins and some PARP family members can subsequently add additional ADP-ribose units to generate a PAR chain. The hydrolysis of PAR chains is catalysed by PAR glycohydrolase (PARG). PARG is unable to cleave the mono(ADP-ribose) unit directly linked to the protein and although the enzymatic activity that catalyses this reaction has been detected in mammalian cell extracts, the protein(s) responsible remain unknown. Here, we report the homozygous mutation of the c6orf130 gene in patients with severe neurodegeneration, and identify C6orf130 as a PARP-interacting protein that removes mono(ADP-ribosyl)ation on glutamate amino acid residues in PARP-modified proteins. X-ray structures and biochemical analysis of C6orf130 suggest a mechanism of catalytic reversal involving a transient C6orf130 lysyl-(ADP-ribose) intermediate. Furthermore, depletion of C6orf130 protein in cells leads to proliferation and DNA repair defects. Collectively, our data suggest that C6orf130 enzymatic activity has a role in the turnover and recycling of protein ADP-ribosylation, and we have implicated the importance of this protein in supporting normal cellular function in humans.
    MeSH term(s) Amino Acid Sequence ; Base Sequence ; Cells, Cultured ; Child ; Child, Preschool ; Family ; Female ; Glycoside Hydrolases/genetics ; Glycoside Hydrolases/physiology ; HEK293 Cells ; HeLa Cells ; Humans ; Male ; Models, Molecular ; Molecular Sequence Data ; Neurodegenerative Diseases/enzymology ; Neurodegenerative Diseases/genetics ; Pedigree ; Poly Adenosine Diphosphate Ribose/genetics ; Poly Adenosine Diphosphate Ribose/physiology ; Protein Processing, Post-Translational/genetics ; Sequence Homology, Amino Acid ; Thiolester Hydrolases/genetics ; Thiolester Hydrolases/physiology
    Chemical Substances Poly Adenosine Diphosphate Ribose (26656-46-2) ; OARD1 protein, human (EC 3.1.2.-) ; Thiolester Hydrolases (EC 3.1.2.-) ; Glycoside Hydrolases (EC 3.2.1.-)
    Language English
    Publishing date 2013-03-12
    Publishing country England
    Document type Case Reports ; Journal Article ; Research Support, N.I.H., Intramural ; Research Support, Non-U.S. Gov't
    ZDB-ID 586044-1
    ISSN 1460-2075 ; 0261-4189
    ISSN (online) 1460-2075
    ISSN 0261-4189
    DOI 10.1038/emboj.2013.51
    Database MEDical Literature Analysis and Retrieval System OnLINE

    More links

    Kategorien

To top