Search

Multimodal foundation models, such as Gemini and ChatGPT, have revolutionized human-machine interactions by seamlessly integrating various forms of data. Developing a universal spoken language model that comprehends a wide range of natural language instructions is critical for bridging communication gaps and facilitating more intuitive interactions. However, the absence of a comprehensive evaluation benchmark poses a significant challenge. We present Dynamic-SUPERB Phase-2, an open and evolving benchmark for the comprehensive evaluation of instruction-based universal speech models. Building upon the first generation, this second version incorporates 125 new tasks contributed collaboratively by the global research community, expanding the benchmark to a total of 180 tasks, making it the largest benchmark for speech and audio evaluation. While the first generation of Dynamic-SUPERB was limited to classification tasks, Dynamic-SUPERB Phase-2 broadens its evaluation capabilities by introducing a wide array of novel and diverse tasks, including regression and sequence generation, across speech, music, and environmental audio. Evaluation results indicate that none of the models performed well universally. SALMONN-13B excelled in English ASR, while WavLLM demonstrated high accuracy in emotion recognition, but current models still require further innovations to handle a broader range of tasks. We will soon open-source all task data and the evaluation pipeline.

Northern auroral regions of Earth were imaged with energetic photons in the 0.1-10 keV range using the High-Resolution Camera (HRC-I) aboard the Chandra X-ray Observatory at 10 epochs (each ~20 min duration) between mid-December 2003 and mid-April 2004. These observations aimed at searching for Earth's soft (<2 keV) X-ray aurora in a comparative study with Jupiter's X-ray aurora, where a pulsating X-ray "hot-spot" has been previously observed by Chandra. The first Chandra soft X-ray observations of Earth's aurora show that it is highly variable (intense arcs, multiple arcs, diffuse patches, at times absent). In at least one of the observations an isolated blob of emission is observed near the expected cusp location. A fortuitous overflight of DMSP satellite F13 provided SSJ/4 energetic particle measurements above a bright arc seen by Chandra on 24 January 2004, 20:01-20:22 UT. A model of the emissions expected strongly suggests that the observed soft X-ray signal is bremsstrahlung and characteristic K-shell line emissions of nitrogen and oxygen in the atmosphere produced by electrons., Comment: See press release on this paper at http://chandra.harvard.edu/press/05_releases/press_122805.html. Journal of Atmospheric and Solar-Terrestrial Physics, in press (2006)

Two experiments investigated preschoolers' acquisition of spatial words in Mandarin Chinese. In one experiment, 5 groups of 10 children at 34, 39, 46, 52, and 57 months were tested for comprehension and production of 14 pairs of Chinese spatial words. In the comprehension test the children were asked to point to pictures corresponding to the words spoken; in the production test they were asked to say the opposite of the word spoken by the experimenter. Results indicated a high degree of consistency in the order of acquisition, which paralleled findings in English acquisition studies and was consistent with theory. However, the theory that unmarked words should be learned before the marked words was not supported. In the second experiment 3 groups of 10 children aged 34, 46, and 57 months were tested for their comprehension of two pairs of spatial terms. An object was placed in relation to another object having one or the other of the following characteristics: (1) inherent top, bottom, front, and back; (2) inherent front/back but not top/bottom; (3) inherent top/bottom but not front/back; and (4) neither top/bottom nor front/back. Results indicated that under all conditions even the youngest children had almost perfect comprehension of top and bottom but imperfect comprehension of front/back for objects without inherent front/back markings. (MSE)

Northern auroral regions of Earth were imaged with energetic photons in the 0.1-10keV range using the High-Resolution Camera (HRC-I) aboard the Chandra X-ray Observatory at 10 epochs (each approx.20 min duration) between mid- December 2003 and mid-April 2004. These observations aimed at searching for Earth's soft (< 2 keV) X-ray aurora in a comparative study with Jupiter's X-ray aurora, where a pulsating X-ray "hot-spot" has been previously observed by Chandra. The first Chandra soft X-ray observations of Earth's aurora show that it is highly variable 0ntense arcs, multiple arcs, diffuse patches, at times absent). In at least one of the observations an isolated blob of emission is observed near the expected cusp location. A fortuitous overflight of DMSP satellite F13 provided SSJ/4 energetic particle measurements above a bright arc seen by Chandra on 24 January 2004, 20:01-20:22 UT. A model of the emissions expected strongly suggests that the observed soft X-ray signal is bremsstrahlung and characteristic K-shell line emissions of nitrogen and oxygen in the atmosphere produced by electrons.

Northern polar "auroral" regions of Earth was observed by High-Resolution Camera in imaging mode (T32C-I) aboard Chandra X-Ray Observatory (CXO) during mid December 2003 - mid April 2004. Ten CXO observations, each approximately 20 min duration, were made in a non-conventional method (due to CXO technical issues), such that Chandra was aimed at a fixed point in sky and the Earth's polar cusp was allowed to drift through the HRC-I field-of-view. The observations were performed when CXO was near apogee and timed during northern winter mostly near midnight (6 hr), except two observations which occurred around 1200 UT, so that northern polar region is entirely in dark and solar fluoresced x-ray contamination can be avoided. These observations were aimed at searching the Earth's soft x-ray aurora and to do a comparative study with Jupiter's x-ray aurora, where a pulsating x-ray hot-spot near the northern magnetic pole has been observed by Chandra that implies a particle source region near Jupiter's magnetopause, and entry of heavy solar wind ions due to high-latitude reconnection as a viable explanation for the soft x-ray emissions. The first Chandra soft (0.1-2 keV) x-ray observations of Earth's aurora show that it is highly variable (intense arc, multiple arcs, diffuse, at times almost absent). In at least one of the observations an isolated blob of emission is observed where we expect cusp to be: giving indication of solar wind charge-exchange signature in x-rays. We are comparing the Chandra x-ray observations with observations at other wavelengths and particle data from Earth-orbiting satellites and solar wind measurements from near-Earth ACE and SOH0 spacecraft. Preliminary results from these unique CXO-Earth observations will be presented and discussed.

Northern auroral regions of Earth were imaged using the High-Resolution Camera (HRC-1) aboard the Chandra X-Ray Observatory (CXO) at 10 epochs (each approx.20 min duration) between mid-December 2003 and mid-April 2004. These observations aimed at searching for Earth s soft (<2 keV) x-ray aurora in a comparative study with Jupiter s x-ray aurora, where a pulsating x-ray "hot-spot" has been previously observed by Chandra. The first Chandra soft x-ray observations of Earth s aurora show that it is highly variable (intense arcs, multiple arcs, diffuse patches, at times absent). In at least one of the observations an isolated blob of emission is observed near the expected cusp location. A fortuitous overflight of DMSP satellite F13 provided SSJ/4 energetic particle measurements above a bright arc seen by Chandra on 24 January 2004, 20:01-20:22 UT. A model of the emissions expected strongly suggests that the observed soft x-ray signal is produced by electron bremsstrahlung.