Audio comprehension—including speech, non-speech sounds, and music—is essential for achieving human-level intelligence. Consequently, AI agents must demonstrate holistic audio understanding to qualify as generally intelligent. However, evaluating auditory intelligence comprehensively remains challenging. To address this gap, we introduce MMAU-Pro, the most comprehensive and rigorously curated benchmark for assessing audio intelligence in AI systems. MMAU-Pro contains 5,305 instances, where each instance has one or more audios paired with human expert-generated question-answer pairs, spanning speech, sound, music, and their combinations. Unlike existing benchmarks, MMAU-Pro evaluates auditory intelligence across 49 unique skills and multiple complex dimensions, including long-form audio comprehension, spatial audio reasoning, multi-audio understanding, among others. All questions are meticulously designed to require deliberate multi-hop reasoning, including both multiple-choice and open-ended response formats. Importantly, audio data is sourced directly ``from the wild" rather than from existing datasets with known distributions. We evaluate 22 leading open-source and proprietary multimodal AI models, revealing significant limitations: even state-of-the-art models such as Gemini 2.5 Flash and Audio Flamingo 3 achieve only 57.33% and 45.9% accuracy, respectively, approaching random performance in multiple categories. Our extensive analysis highlights specific shortcomings and provides novel insights, offering actionable perspectives for the community to enhance future AI systems' progression toward audio general intelligence.
We experimentally elicit views of what exploitation is from over 2,000 subjects. Our experimental design does not test existing theories of exploitation. Rather, it focuses on more fundamental properties that are the building blocks for these theories. We find, first, that exploitation is not a vacuous concept: Not all economic interactions are deemed exploitative. Second, contrary to several of the major approaches in the literature, both inequalities in the distribution of economic gains and asymmetric power relations contribute to exploitative relations. What matters most is the interaction of power and inequality: The effect of both elements together is significantly greater than the sum of each on their own. Finally, and perhaps remarkably, we found no major differences in exploitation ascriptions between experts and lay subjects. These findings have implications for the ethics of employment contracts, particularly in the context of sweatshop labor.
Past research has often found that children appraise hypothetical gender-nonconforming (GNC) peers less positively, but the role of children's real-life familiarity with GNC peers has seldom been explored in relation to such appraisals. Past research on intergroup contact suggests that real-life familiarity would have a positive influence on intergroup attitudes. In the present study, this hypothesis was tested by examining how 8- to 11-year-olds' (N = 186) real-life familiarity with trans and gender-nonconforming (TGNC) children related to their trait perception ratings (e.g., nice, smart, friendly) of hypothetical GNC peers. In alignment with the hypothesis, child participants who were more familiar with TGNC children in real-life rated hypothetical GNC targets more positively. Also, child participants who were more familiar with TGNC children rated GC targets less positively. These findings have important implications for parents, teachers, and researchers who are developing and implementing interventions aimed at improving children's attitudes toward GNC peers.
Studying the remnants of young core-collapse supernovae (SNe) can yield insight into the chemical composition of their progenitors and the geometry of the explosions. The supernova remnant (SNR) G292.0+1.8 is one of only three known oxygen-rich SNRs in the galaxy-remnants of core collapse of which relatively pure fragments of ejecta can be seen. Several dozen ejecta knots from G292.0+1.8 were the subject of a proper motion analysis, based on [O iii] 5007 & Aring; images taken over a 22 yr baseline by P. F. Winkler et al. They determined that the transverse velocities of the filaments are linearly proportional to their distances from a common expansion center; thus, the O-rich filaments have been traveling with little deceleration since the initial SN event, similar to 3000 yr ago. In this paper, we use optical spectra of G292.0+1.8, all taken from the Cerro Tololo Inter-American Observatory, to measure radial velocities for 93 knots. Assuming undecelerated expansion, as indicated by the proper motions, the radial velocity should be proportional to the distance from the center along the line of sight, just as the proper motions are proportional to the transverse distance. Therefore, we can map the three-dimensional structure and kinematics of the SNR. We find that the knots generally follow a broad biconical distribution, suggesting that the SN explosion produced broad jets of ejecta. This structure is similar to that seen in some other young core-collapse SNRs.
A graph G is universal for a (finite) family H of graphs if every H is an element of H is a subgraph of G. For a given family 9-l, the goal is to determine the smallest number of edges an 9-l-universal graph can have. With the aim of unifying a number of recent results, we consider a family of graphs with bounded density. In particular, we construct a graph with O-d (n(2-1/(d+1))) edges which contains every n-vertex graph with density at most d is an element of Q (d >= 1), which is close to a Omega(n(2-1/d)) lower bound obtained by counting lifts of a carefully chosen (small) graph. When restricting the maximum degree of such graphs to be constant, we obtain near-optimal universality. If we further assume d is an element of N, we get an asymptotically optimal construction.
