Evaluating Generative AI Systems is a Social Science Measurement Challenge
Hanna Wallach
Meera Desai
Nicholas Pangakis
A. Feder Cooper
Angelina Wang
Solon Barocas
Alexandra Chouldechova
Chad Atalla
Su Lin Blodgett
Emily Corvi
P. A. Dow
Jean Garcia-Gathright
Alexandra Olteanu
Stefanie Reed
Emily Sheng
Dan Vann
Jennifer Wortman Vaughan
Matthew Vogel
Hannah Washington
Abigail Z. Jacobs … (see 1 more)
Microsoft Research
Across academia, industry, and government, there is an increasing awareness that the measurement tasks involved in evaluating generative AI … (see more)(GenAI) systems are especially difficult. We argue that these measurement tasks are highly reminiscent of measurement tasks found throughout the social sciences. With this in mind, we present a framework, grounded in measurement theory from the social sciences, for measuring concepts related to the capabilities, impacts, opportunities, and risks of GenAI systems. The framework distinguishes between four levels: the background concept, the systematized concept, the measurement instrument(s), and the instance-level measurements themselves. This four-level approach differs from the way measurement is typically done in ML, where researchers and practitioners appear to jump straight from background concepts to measurement instruments, with little to no explicit systematization in between. As well as surfacing assumptions, thereby making it easier to understand exactly what the resulting measurements do and do not mean, this framework has two important implications for evaluating evaluations: First, it can enable stakeholders from different worlds to participate in conceptual debates, broadening the expertise involved in evaluating GenAI systems. Second, it brings rigor to operational debates by offering a set of lenses for interrogating the validity of measurement instruments and their resulting measurements.
2024-11-01
arXiv (published)
doi.org
arxiv.org
GitChameleon: Unmasking the Version-Switching Capabilities of Code Generation Models
Nizar Islah
Justine Gehring
Diganta Misra
Eilif Muller
Irina Rish
Terry Yue Zhuo
Massimo Caccia
2024-11-01
arXiv (published)
doi.org
arxiv.org
Imagining and building wise machines: The centrality of AI metacognition
Samuel G. B. Johnson
Amir-Hossein Karimi
Yoshua Bengio
Nick Chater
Tobias Gerstenberg
Kate Larson
Sydney Levine
Melanie Mitchell
Iyad Rahwan
Bernhard Schölkopf
Igor Grossmann
2024-11-01
arXiv (published)
doi.org
arxiv.org
Impact of LLM-based Review Comment Generation in Practice: A Mixed Open-/Closed-source User Study
Doriane Olewicki
Leuson Da Silva
Suhaib Mujahid
Arezou Amini
Benjamin Mah
Marco Castelluccio
Sarra Habchi
Foutse Khomh
Bram Adams
2024-11-01
arXiv (published)
doi.org
arxiv.org
Predictive Modeling of Body Image Dissatisfaction in People With Type 1 Diabetes
COURTNEY SOUTH
SHAHRYAR EBRAHIMI
A. Brazeau
Maria Cutumisu
2024-11-01
Canadian Journal of Diabetes (published)
doi.org
Predictive Modeling of Body Image Dissatisfaction in People With Type 1 Diabetes
COURTNEY SOUTH
SHAHRYAR EBRAHIMI
ANNE-SOPHIE BRAZEAU
Maria Cutumisu
2024-11-01
Canadian Journal of Diabetes (published)
doi.org
A protocol for trustworthy EEG decoding with neural networks
Davide Borra
Elisa Magosso
Mirco Ravanelli
2024-11-01
Neural Networks (published)
doi.org
Soft Condorcet Optimization for Ranking of General Agents
Marc Lanctot
Kate Larson
Michael Kaisers
Quentin Berthet
Ian Gemp
Manfred Diaz
Roberto-Rafael Maura-Rivero
Yoram Bachrach
Anna Koop
Doina Precup
A common way to drive progress of AI models and agents is to compare their performance on standardized benchmarks. Comparing the performance… (see more) of general agents requires aggregating their individual performances across a potentially wide variety of different tasks. In this paper, we describe a novel ranking scheme inspired by social choice frameworks, called Soft Condorcet Optimization (SCO), to compute the optimal ranking of agents: the one that makes the fewest mistakes in predicting the agent comparisons in the evaluation data. This optimal ranking is the maximum likelihood estimate when evaluation data (which we view as votes) are interpreted as noisy samples from a ground truth ranking, a solution to Condorcet's original voting system criteria. SCO ratings are maximal for Condorcet winners when they exist, which we show is not necessarily true for the classical rating system Elo. We propose three optimization algorithms to compute SCO ratings and evaluate their empirical performance. When serving as an approximation to the Kemeny-Young voting method, SCO rankings are on average 0 to 0.043 away from the optimal ranking in normalized Kendall-tau distance across 865 preference profiles from the PrefLib open ranking archive. In a simulated noisy tournament setting, SCO achieves accurate approximations to the ground truth ranking and the best among several baselines when 59\% or more of the preference data is missing. Finally, SCO ranking provides the best approximation to the optimal ranking, measured on held-out test sets, in a problem containing 52,958 human players across 31,049 games of the classic seven-player game of Diplomacy.
2024-11-01
arXiv (published)
doi.org
arxiv.org
SpeechBrain-MOABB: An open-source Python library for benchmarking deep neural networks applied to EEG signals
Davide Borra
Francesco Paissan
Mirco Ravanelli
2024-11-01
Computers in Biology and Medicine (published)
doi.org
Tracing Optimization for Performance Modeling and Regression Detection
Kaveh Shahedi
Heng Li
Maxime Lamothe
Foutse Khomh
Software performance modeling plays a crucial role in developing and maintaining software systems. A performance model analytically describe… (see more)s the relationship between the performance of a system and its runtime activities. This process typically examines various aspects of a system's runtime behavior, such as the execution frequency of functions or methods, to forecast performance metrics like program execution time. By using performance models, developers can predict expected performance and thereby effectively identify and address unexpected performance regressions when actual performance deviates from the model's predictions. One common and precise method for capturing performance behavior is software tracing, which involves instrumenting the execution of a program, either at the kernel level (e.g., system calls) or application level (e.g., function calls). However, due to the nature of tracing, it can be highly resource-intensive, making it impractical for production environments where resources are limited. In this work, we propose statistical approaches to reduce tracing overhead by identifying and excluding performance-insensitive code regions, particularly application-level functions, from tracing while still building accurate performance models that can capture performance degradations. By selecting an optimal set of functions to be traced, we can construct optimized performance models that achieve an R-2 score of up to 99% and, sometimes, outperform full tracing models (models using non-optimized tracing data), while significantly reducing the tracing overhead by more than 80% in most cases. Our optimized performance models can also capture performance regressions in our studied programs effectively, demonstrating their usefulness in real-world scenarios. Our approach is fully automated, making it ready to be used in production environments with minimal human effort.
2024-11-01
arXiv (published)
doi.org
arxiv.org
Unsupervised Object Discovery: A Comprehensive Survey and Unified Taxonomy
Jos'e-Fabian Villa-V'asquez
Marco Pedersoli
Unsupervised object discovery is commonly interpreted as the task of localizing and/or categorizing objects in visual data without the need … (see more)for labeled examples. While current object recognition methods have proven highly effective for practical applications, the ongoing demand for annotated data in real-world scenarios drives research into unsupervised approaches. Furthermore, existing literature in object discovery is both extensive and diverse, posing a significant challenge for researchers that aim to navigate and synthesize this knowledge. Motivated by the evidenced interest in this avenue of research, and the lack of comprehensive studies that could facilitate a holistic understanding of unsupervised object discovery, this survey conducts an in-depth exploration of the existing approaches and systematically categorizes this compendium based on the tasks addressed and the families of techniques employed. Additionally, we present an overview of common datasets and metrics, highlighting the challenges of comparing methods due to varying evaluation protocols. This work intends to provide practitioners with an insightful perspective on the domain, with the hope of inspiring new ideas and fostering a deeper understanding of object discovery approaches.
2024-11-01
arXiv (published)
doi.org
arxiv.org
Words Matter: Leveraging Individual Text Embeddings for Code Generation in CLIP Test-Time Adaptation
Shambhavi Mishra
Julio Silva-Rodríguez
Ismail Ben Ayed
Marco Pedersoli
Jose Dolz
Vision-language foundation models, such as CLIP, have shown unprecedented zero-shot performance across a wide range of tasks. Nevertheless, … (see more)these models may be unreliable under distributional shifts, as their performance is significantly degraded. In this work, we explore how to efficiently leverage class text information to mitigate these distribution drifts encountered by large pre-trained vision-language models (VLMs) during test-time inference. In particular, we propose to generate pseudo-labels for the test-time samples by exploiting generic class text embeddings as fixed centroids of a label assignment problem, which is efficiently solved with Optimal Transport. Furthermore, the proposed adaptation method (CLIP-OT) integrates a multiple template knowledge distillation approach, which replicates multi-view contrastive learning strategies in unsupervised representation learning but without incurring additional computational complexity. Extensive experiments on multiple popular test-time adaptation benchmarks presenting diverse complexity empirically show the superiority of CLIP-OT, achieving performance gains of up to 7% over recent state-of-the-art methods, yet being computationally and memory efficient.
2024-11-01
arXiv (published)
doi.org
arxiv.org