LIVIVO - Das Suchportal für Lebenswissenschaften

switch to English language
Erweiterte Suche

Suchergebnis

Treffer 1 - 3 von insgesamt 3

Suchoptionen

  1. Buch ; Online: GPT is becoming a Turing machine

    Jojic, Ana / Wang, Zhen / Jojic, Nebojsa

    Here are some ways to program it

    2023  

    Abstract: We demonstrate that, through appropriate prompting, GPT-3 family of models can be triggered to perform iterative behaviours necessary to execute (rather than just write or recall) programs that involve loops, including several popular algorithms found in ...

    Abstract We demonstrate that, through appropriate prompting, GPT-3 family of models can be triggered to perform iterative behaviours necessary to execute (rather than just write or recall) programs that involve loops, including several popular algorithms found in computer science curricula or software developer interviews. We trigger execution and description of Iterations by Regimenting Self-Attention (IRSA) in one (or a combination) of three ways: 1) Using strong repetitive structure in an example of an execution path of a target program for one particular input, 2) Prompting with fragments of execution paths, and 3) Explicitly forbidding (skipping) self-attention to parts of the generated text. On a dynamic program execution, IRSA leads to larger accuracy gains than replacing the model with the much more powerful GPT-4. IRSA has promising applications in education, as the prompts and responses resemble student assignments in data structures and algorithms classes. Our findings hold implications for evaluating LLMs, which typically target the in-context learning: We show that prompts that may not even cover one full task example can trigger algorithmic behaviour, allowing solving problems previously thought of as hard for LLMs, such as logical puzzles. Consequently, prompt design plays an even more critical role in LLM performance than previously recognized.

    Comment: 25 pages, 1 figure
    Schlagwörter Computer Science - Computation and Language
    Thema/Rubrik (Code) 004
    Erscheinungsdatum 2023-03-24
    Erscheinungsland us
    Dokumenttyp Buch ; Online
    Datenquelle BASE - Bielefeld Academic Search Engine (Lebenswissenschaftliche Auswahl)

    Zusatzmaterialien

    Kategorien

  2. Buch ; Online: Resolving label uncertainty with implicit posterior models

    Rolf, Esther / Malkin, Nikolay / Graikos, Alexandros / Jojic, Ana / Robinson, Caleb / Jojic, Nebojsa

    2022  

    Abstract: We propose a method for jointly inferring labels across a collection of data samples, where each sample consists of an observation and a prior belief about the label. By implicitly assuming the existence of a generative model for which a differentiable ... ...

    Abstract We propose a method for jointly inferring labels across a collection of data samples, where each sample consists of an observation and a prior belief about the label. By implicitly assuming the existence of a generative model for which a differentiable predictor is the posterior, we derive a training objective that allows learning under weak beliefs. This formulation unifies various machine learning settings; the weak beliefs can come in the form of noisy or incomplete labels, likelihoods given by a different prediction mechanism on auxiliary input, or common-sense priors reflecting knowledge about the structure of the problem at hand. We demonstrate the proposed algorithms on diverse problems: classification with negative training examples, learning from rankings, weakly and self-supervised aerial imagery segmentation, co-segmentation of video frames, and coarsely supervised text classification.

    Comment: UAI 2022; code: https://github.com/estherrolf/implicit-posterior
    Schlagwörter Computer Science - Machine Learning ; Statistics - Machine Learning
    Thema/Rubrik (Code) 006
    Erscheinungsdatum 2022-02-28
    Erscheinungsland us
    Dokumenttyp Buch ; Online
    Datenquelle BASE - Bielefeld Academic Search Engine (Lebenswissenschaftliche Auswahl)

    Zusatzmaterialien

    Kategorien

  3. Artikel ; Online: Circulating cancer-specific CD8 T cell frequency is associated with response to PD-1 blockade in Merkel cell carcinoma.

    Pulliam, Thomas / Jani, Saumya / Jing, Lichen / Ryu, Heeju / Jojic, Ana / Shasha, Carolyn / Zhang, Jiajia / Kulikauskas, Rima / Church, Candice / Garnett-Benson, Charlie / Gooley, Ted / Chapuis, Aude / Paulson, Kelly / Smith, Kellie N / Pardoll, Drew M / Newell, Evan W / Koelle, David M / Topalian, Suzanne L / Nghiem, Paul

    Cell reports. Medicine

    2024  Band 5, Heft 2, Seite(n) 101412

    Abstract: Understanding cancer immunobiology has been hampered by difficulty identifying cancer-specific T cells. Merkel cell polyomavirus (MCPyV) causes most Merkel cell carcinomas (MCCs). All patients with virus-driven MCC express MCPyV oncoproteins, ... ...

    Abstract Understanding cancer immunobiology has been hampered by difficulty identifying cancer-specific T cells. Merkel cell polyomavirus (MCPyV) causes most Merkel cell carcinomas (MCCs). All patients with virus-driven MCC express MCPyV oncoproteins, facilitating identification of virus (cancer)-specific T cells. We studied MCPyV-specific T cells from 27 patients with MCC using MCPyV peptide-HLA-I multimers, 26-color flow cytometry, single-cell transcriptomics, and T cell receptor (TCR) sequencing. In a prospective clinical trial, higher circulating MCPyV-specific CD8 T cell frequency before anti-PD-1 treatment was strongly associated with 2-year recurrence-free survival (75% if detectable, 0% if undetectable, p = 0.0018; ClinicalTrial.gov: NCT02488759). Intratumorally, such T cells were typically present, but their frequency did not significantly associate with response. Circulating MCPyV-specific CD8 T cells had increased stem/memory and decreased exhaustion signatures relative to their intratumoral counterparts. These results suggest that cancer-specific CD8 T cells in the blood may play a role in anti-PD-1 responses. Thus, strategies that augment their number or mobilize them into tumors could improve outcomes.
    Mesh-Begriff(e) Humans ; Carcinoma, Merkel Cell/drug therapy ; Carcinoma, Merkel Cell/pathology ; CD8-Positive T-Lymphocytes/pathology ; Programmed Cell Death 1 Receptor ; Prospective Studies ; Skin Neoplasms/drug therapy ; Skin Neoplasms/pathology ; Clinical Trials as Topic
    Chemische Substanzen Programmed Cell Death 1 Receptor
    Sprache Englisch
    Erscheinungsdatum 2024-02-10
    Erscheinungsland United States
    Dokumenttyp Journal Article
    ISSN 2666-3791
    ISSN (online) 2666-3791
    DOI 10.1016/j.xcrm.2024.101412
    Datenquelle MEDical Literature Analysis and Retrieval System OnLINE

    Zusatzmaterialien

    Kategorien

Zum Seitenanfang