LIVIVO - The Search Portal for Life Sciences

zur deutschen Oberfläche wechseln
Advanced search

Search results

Result 1 - 10 of total 22

Search options

  1. Book ; Online: Decision-Oriented Dialogue for Human-AI Collaboration

    Lin, Jessy / Tomlin, Nicholas / Andreas, Jacob / Eisner, Jason

    2023  

    Abstract: We describe a class of tasks called decision-oriented dialogues, in which AI assistants must collaborate with one or more humans via natural language to help them make complex decisions. We formalize three domains in which users face everyday decisions: ( ...

    Abstract We describe a class of tasks called decision-oriented dialogues, in which AI assistants must collaborate with one or more humans via natural language to help them make complex decisions. We formalize three domains in which users face everyday decisions: (1) choosing an assignment of reviewers to conference papers, (2) planning a multi-step itinerary in a city, and (3) negotiating travel plans for a group of friends. In each of these settings, AI assistants and users have disparate abilities that they must combine to arrive at the best decision: assistants can access and process large amounts of information, while users have preferences and constraints external to the system. For each task, we build a dialogue environment where agents receive a reward based on the quality of the final decision they reach. Using these environments, we collect human-human dialogues with humans playing the role of assistant. To compare how current AI assistants communicate in these settings, we present baselines using large language models in self-play. Finally, we highlight a number of challenges models face in decision-oriented dialogues, ranging from efficient communication to reasoning and optimization, and release our environments as a testbed for future modeling work.
    Keywords Computer Science - Computation and Language ; Computer Science - Artificial Intelligence
    Subject code 004 ; 006
    Publishing date 2023-05-31
    Publishing country us
    Document type Book ; Online
    Database BASE - Bielefeld Academic Search Engine (life sciences selection)

    More links

    Kategorien

  2. Book ; Online: Revisiting Entropy Rate Constancy in Text

    Verma, Vivek / Tomlin, Nicholas / Klein, Dan

    2023  

    Abstract: The uniform information density (UID) hypothesis states that humans tend to distribute information roughly evenly across an utterance or discourse. Early evidence in support of the UID hypothesis came from Genzel & Charniak (2002), which proposed an ... ...

    Abstract The uniform information density (UID) hypothesis states that humans tend to distribute information roughly evenly across an utterance or discourse. Early evidence in support of the UID hypothesis came from Genzel & Charniak (2002), which proposed an entropy rate constancy principle based on the probability of English text under n-gram language models. We re-evaluate the claims of Genzel & Charniak (2002) with neural language models, failing to find clear evidence in support of entropy rate constancy. We conduct a range of experiments across datasets, model sizes, and languages and discuss implications for the uniform information density hypothesis and linguistic theories of efficient communication more broadly.

    Comment: Findings of EMNLP 2023
    Keywords Computer Science - Computation and Language
    Publishing date 2023-05-19
    Publishing country us
    Document type Book ; Online
    Database BASE - Bielefeld Academic Search Engine (life sciences selection)

    More links

    Kategorien

  3. Book ; Online: Ghostbuster

    Verma, Vivek / Fleisig, Eve / Tomlin, Nicholas / Klein, Dan

    Detecting Text Ghostwritten by Large Language Models

    2023  

    Abstract: We introduce Ghostbuster, a state-of-the-art system for detecting AI-generated text. Our method works by passing documents through a series of weaker language models, running a structured search over possible combinations of their features, and then ... ...

    Abstract We introduce Ghostbuster, a state-of-the-art system for detecting AI-generated text. Our method works by passing documents through a series of weaker language models, running a structured search over possible combinations of their features, and then training a classifier on the selected features to predict whether documents are AI-generated. Crucially, Ghostbuster does not require access to token probabilities from the target model, making it useful for detecting text generated by black-box models or unknown model versions. In conjunction with our model, we release three new datasets of human- and AI-generated text as detection benchmarks in the domains of student essays, creative writing, and news articles. We compare Ghostbuster to a variety of existing detectors, including DetectGPT and GPTZero, as well as a new RoBERTa baseline. Ghostbuster achieves 99.0 F1 when evaluated across domains, which is 5.9 F1 higher than the best preexisting model. It also outperforms all previous approaches in generalization across writing domains (+7.5 F1), prompting strategies (+2.1 F1), and language models (+4.4 F1). We also analyze the robustness of our system to a variety of perturbations and paraphrasing attacks and evaluate its performance on documents written by non-native English speakers.
    Keywords Computer Science - Computation and Language ; Computer Science - Artificial Intelligence
    Subject code 006
    Publishing date 2023-05-24
    Publishing country us
    Document type Book ; Online
    Database BASE - Bielefeld Academic Search Engine (life sciences selection)

    More links

    Kategorien

  4. Book ; Online: Understanding Game-Playing Agents with Natural Language Annotations

    Tomlin, Nicholas / He, Andre / Klein, Dan

    2022  

    Abstract: We present a new dataset containing 10K human-annotated games of Go and show how these natural language annotations can be used as a tool for model interpretability. Given a board state and its associated comment, our approach uses linear probing to ... ...

    Abstract We present a new dataset containing 10K human-annotated games of Go and show how these natural language annotations can be used as a tool for model interpretability. Given a board state and its associated comment, our approach uses linear probing to predict mentions of domain-specific terms (e.g., ko, atari) from the intermediate state representations of game-playing agents like AlphaGo Zero. We find these game concepts are nontrivially encoded in two distinct policy networks, one trained via imitation learning and another trained via reinforcement learning. Furthermore, mentions of domain-specific terms are most easily predicted from the later layers of both models, suggesting that these policy networks encode high-level abstractions similar to those used in the natural language annotations.
    Keywords Computer Science - Computation and Language ; Computer Science - Artificial Intelligence
    Publishing date 2022-04-15
    Publishing country us
    Document type Book ; Online
    Database BASE - Bielefeld Academic Search Engine (life sciences selection)

    More links

    Kategorien

  5. Article: Exploring Student Experiences of an Undergraduate Certificate in Veterinary Medical Education.

    Hughes, Kirsty / Hudson, Neil / Bell, Catriona / Tomlin, Nina / Handel, Ian / Chinnery, Sarah / Rhind, Susan

    Journal of veterinary medical education

    2022  , Page(s) e20210098

    Abstract: The ability to teach is recognized as a core skill for many professionals, including veterinarians, but undergraduate opportunities to develop this skill are not always available. A complementary teaching certificate offered during the clinical years of ... ...

    Abstract The ability to teach is recognized as a core skill for many professionals, including veterinarians, but undergraduate opportunities to develop this skill are not always available. A complementary teaching certificate offered during the clinical years of an undergraduate veterinary program was evaluated to investigate student experiences of the program and the perceived benefits and challenges of participating. The study used a mixed methods approach with questionnaires to provide an overview of the participant experience and semi-structured interviews to gain a deeper insight into students' experiences of the program. Two cohorts completed questionnaires comprising Likert-style and open-ended questions on the 3-year teaching certificate, the first cohort after 1 year of the program and the second cohort at completion. Interviews with participants from both cohorts were thematically analyzed to identify recurring themes. An average of 27% of students per academic year enrolled in the certificate program, most of whom completed it. Additionally, four to six students per cohort applied for Associate Fellow of the Higher Education Academy (AFHEA), and 19 students have achieved this recognition. Key themes from the data included that students felt the certificate built their confidence, increased their veterinary knowledge, and helped them become better teachers, with time management and reflection the biggest challenges. The Undergraduate Certificate of Veterinary Medical Education was seen as a good teaching foundation, while working toward the AFHEA provided some insight into higher education and academic careers. A structured teaching program offers students the opportunity to develop their learning and reflection both as students and future educators.
    Language English
    Publishing date 2022-05-17
    Publishing country Canada
    Document type Journal Article
    ISSN 0748-321X
    ISSN 0748-321X
    DOI 10.3138/jvme-2021-0098
    Database MEDical Literature Analysis and Retrieval System OnLINE

    More links

    Kategorien

  6. Article: Generalized Electrical Substitution Methods and Detectors for Absolute Optical Power Measurements.

    Woods, S I / Neira, J E / Proctor, J E / Rice, J P / Tomlin, N A / White, M G / Stephens, M S / Lehman, J H

    Metrologia

    2023  Volume 59, Issue 4

    Abstract: We have developed generalized methods for electrical substitution optical measurements, as well as cryogenic detectors which can be used to implement them. The new methods detailed here enable measurement of arbitrary periodic waveforms by an electrical ... ...

    Abstract We have developed generalized methods for electrical substitution optical measurements, as well as cryogenic detectors which can be used to implement them. The new methods detailed here enable measurement of arbitrary periodic waveforms by an electrical substitution radiometer (ESR), which means that spectral and dynamic optical power can be absolutely calibrated directly by a primary standard detector. Cryogenic ESRs are not often used directly by researchers for optical calibrations due to their slow response times and cumbersome operation. We describe two types of ESRs with fast response times, including newly developed cryogenic bolometers with carbon nanotube absorbers, which are manufacturable by standard microfabrication techniques. These detectors have response times near 10 ms, spectral coverage from the ultraviolet to far-infrared, and are ideal for use with generalized electrical substitution. In our first tests of the generalized electrical substitution method with FTS, we have achieved uncertainty in detector response of 0.13 % (k=1) and total measurement uncertainty of 1.1 % (k=1) in the mid-infrared for spectral detector responsivity calibrations. The generalized method and fast detectors greatly expand the range of optical power calibrations which can be made using a wideband primary standard detector, which can shorten calibration chains and improve uncertainties.
    Language English
    Publishing date 2023-01-17
    Publishing country England
    Document type Journal Article
    ZDB-ID 207790-5
    ISSN 0026-1394
    ISSN 0026-1394
    DOI 10.1088/1681-7575/ac72dc
    Database MEDical Literature Analysis and Retrieval System OnLINE

    More links

    Kategorien

  7. Book ; Online: Pragmatics in Language Grounding

    Fried, Daniel / Tomlin, Nicholas / Hu, Jennifer / Patel, Roma / Nematzadeh, Aida

    Phenomena, Tasks, and Modeling Approaches

    2022  

    Abstract: People rely heavily on context to enrich meaning beyond what is literally said, enabling concise but effective communication. To interact successfully and naturally with people, user-facing artificial intelligence systems will require similar skills in ... ...

    Abstract People rely heavily on context to enrich meaning beyond what is literally said, enabling concise but effective communication. To interact successfully and naturally with people, user-facing artificial intelligence systems will require similar skills in pragmatics: relying on various types of context -- from shared linguistic goals and conventions, to the visual and embodied world -- to use language effectively. We survey existing grounded settings and pragmatic modeling approaches and analyze how the task goals, environmental contexts, and communicative affordances in each work enrich linguistic meaning. We present recommendations for future grounded task design to naturally elicit pragmatic phenomena, and suggest directions that focus on a broader range of communicative contexts and affordances.
    Keywords Computer Science - Computation and Language
    Publishing date 2022-11-15
    Publishing country us
    Document type Book ; Online
    Database BASE - Bielefeld Academic Search Engine (life sciences selection)

    More links

    Kategorien

  8. Book ; Online: Automated Crossword Solving

    Wallace, Eric / Tomlin, Nicholas / Xu, Albert / Yang, Kevin / Pathak, Eshaan / Ginsberg, Matthew / Klein, Dan

    2022  

    Abstract: We present the Berkeley Crossword Solver, a state-of-the-art approach for automatically solving crossword puzzles. Our system works by generating answer candidates for each crossword clue using neural question answering models and then combines loopy ... ...

    Abstract We present the Berkeley Crossword Solver, a state-of-the-art approach for automatically solving crossword puzzles. Our system works by generating answer candidates for each crossword clue using neural question answering models and then combines loopy belief propagation with local search to find full puzzle solutions. Compared to existing approaches, our system improves exact puzzle accuracy from 71% to 82% on crosswords from The New York Times and obtains 99.9% letter accuracy on themeless puzzles. Additionally, in 2021, a hybrid of our system and the existing Dr.Fill system outperformed all human competitors for the first time at the American Crossword Puzzle Tournament. To facilitate research on question answering and crossword solving, we analyze our system's remaining errors and release a dataset of over six million question-answer pairs.

    Comment: ACL 2022
    Keywords Computer Science - Computation and Language
    Publishing date 2022-05-19
    Publishing country us
    Document type Book ; Online
    Database BASE - Bielefeld Academic Search Engine (life sciences selection)

    More links

    Kategorien

  9. Article ; Online: Carbon nanotube electrical-substitution cryogenic radiometer: initial results.

    Tomlin, N A / Lehman, J H

    Optics letters

    2013  Volume 38, Issue 2, Page(s) 175–177

    Abstract: A carbon nanotube cryogenic radiometer (CNCR) has been fabricated for electrical-substitution optical power measurements. The CNCR employs vertically aligned multiwall carbon nanotube arrays (VANTAs) as the absorber, heater, and thermistor, with a ... ...

    Abstract A carbon nanotube cryogenic radiometer (CNCR) has been fabricated for electrical-substitution optical power measurements. The CNCR employs vertically aligned multiwall carbon nanotube arrays (VANTAs) as the absorber, heater, and thermistor, with a micromachined silicon substrate as the weak thermal link. Compared to conventional cryogenic radiometers, the CNCR is simpler, more easily reproduced and disseminated, orders of magnitude faster, and can operate over a wide range of wavelengths without the need for a receiver cavity. We describe initial characterization results of the radiometer at 3.9 K, comparing electrical measurements and fiber-coupled optical measurements from 50 μW to 1.5 mW at the wavelength of 1550 nm. We find the response to input electrical and optical power is equivalent to within our measurement uncertainty, which is currently limited by the experimental setup (large temperature fluctuations of the cold stage) rather than the device itself. With improvements in the temperature stability, the performance of the CNCR should be limited only by our ability to measure the reflectance of the optical absorber VANTA.
    Language English
    Publishing date 2013-01-15
    Publishing country United States
    Document type Journal Article
    ISSN 1539-4794
    ISSN (online) 1539-4794
    DOI 10.1364/OL.38.000175
    Database MEDical Literature Analysis and Retrieval System OnLINE

    More links

    Kategorien

  10. Article ; Online: Micro-DRIFTS for small area hyper-black spectroscopy.

    McArdle, Patrick / Yung, Christopher / Tomlin, Nathan / Lehman, John / Stephens, Michelle

    Optics express

    2019  Volume 31, Issue 26, Page(s) 44328–44339

    Abstract: We have developed a low-cost micro-diffuse reflectance infrared Fourier transform spectroscopic (micro-DRIFTS) setup for measuring the reflectance of small area diffuse samples. The system performance is characterized and then demonstrated on small area ... ...

    Abstract We have developed a low-cost micro-diffuse reflectance infrared Fourier transform spectroscopic (micro-DRIFTS) setup for measuring the reflectance of small area diffuse samples. The system performance is characterized and then demonstrated on small area vertically aligned carbon nanotube (VACNT) samples. We find that our system can measure samples with a spatial resolution of approximately 140 µm with sensitivities of 10s of ppm in the 2 µm - 18 µm spectral window. Our uncertainty budget is presented along with how our measured reflectance can be equated to directional-hemispherical reflectance.
    Language English
    Publishing date 2019-12-20
    Publishing country United States
    Document type Journal Article
    ZDB-ID 1491859-6
    ISSN 1094-4087 ; 1094-4087
    ISSN (online) 1094-4087
    ISSN 1094-4087
    DOI 10.1364/OE.506419
    Database MEDical Literature Analysis and Retrieval System OnLINE

    More links

    Kategorien

To top