Oceania
Predict Responsibly: Improving Fairness and Accuracy by Learning to Defer
David Madras, Toni Pitassi, Richard Zemel
Can humans and machines make decisions jointly?
COLA: Decentralized Linear Learning
Neural Information Processing Systems http://nips.cc/
Pornhub Is Urging Tech Giants to Enact Device-Based Age Verification
The company sent letters to Apple, Google, and Microsoft pushing for an alternative way to keep minors from viewing porn, as US and UK laws have caused its traffic to plummet. In letters sent to Apple, Google, and Microsoft this week, Pornhub's parent company urged the tech giants to support device-based age verification in their app stores and across their operating systems, WIRED has learned. "Based on our real-world experience with existing age assurance laws, we strongly support the initiative to protect minors online," reads the letter sent by Anthony Penhale, chief legal officer for Aylo, which owns Pornhub, Brazzers, Redtube, and YouPorn. "However, we have found site-based age assurance approaches to be fundamentally flawed and counterproductive." The letter adds that site-based age verification methods have "failed to achieve their primary objective: protecting minors from accessing age-inappropriate material online."
VLM4Bio: A Benchmark Dataset to Evaluate Pretrained Vision-Language Models for Trait Discovery from Biological Images Supplementary Materials
Fish-10K was manually curated with the help of biological experts co-authored in this paper.
Identifying Functionally Important Features with End-to-End Sparse Dictionary Learning Dan Braun Jordan Taylor Nicholas Goldowsky-Dill Lee Sharkey
Identifying the features learned by neural networks is a core challenge in mechanistic interpretability. Sparse autoencoders (SAEs), which learn a sparse, overcomplete dictionary that reconstructs a network's internal activations, have been used to identify these features. However, SAEs may learn more about the structure of the dataset than the computational structure of the network. There is therefore only indirect reason to believe that the directions found in these dictionaries are functionally important to the network. We propose end-to-end (e2e) sparse dictionary learning, a method for training SAEs that ensures the features learned are functionally important by minimizing the KL divergence between the output distributions of the original model and the model with SAE activations inserted.
Man Ho Lam 1 Eric John Li
Evaluating Large Language Models' (LLMs) anthropomorphic capabilities has
Frequency-domain MLPs are More Effective Learners in Time Series Forecasting
Then, we propose FreTS, a simple yet effective architecture built upon Fre quency-domain MLPs for T ime S eries forecasting.
FIND: A Function Description Benchmark for Evaluating Interpretability Methods Sarah Schwettmann
The central task of interpretability research is to explain the functions that AI systems learn from data. Investigating these functions requires experimentation with trained models, using tools that incorporate varying degrees of human input. Hand-tooled approaches that rely on close manual inspection [Zeiler and Fergus, 2014, Zhou et al., 2014, Mahendran and V edaldi, 2015, Olah et al., 2017, 2020, Elhage et al., 2021] or search for predefined phenomena [Wang et al., 2022, Nanda