
This paper provides a unified treatment to the recovery of structured signals living in a star-shaped set from general quantized measurements Q(Ax−τ), where A is a sensing matrix, τ is a vector of (possibly random) quantization thresholds, and Q denotes an L-level quantizer. The ideal estimator of hamming distance minimization (HDM) is optimal but typically infeasible to compute. Under sub-Gaussian design, we study the projected gradient descent (PGD) algorithm with respect to the one-sided ℓ1-loss and identify the conditions under which PGD achieves the same error rate as HDM, up to logarithmic factors. These conditions include estimates of the separation probability, small-ball probability and some moment bounds that are easy to validate. For the multi-bit case, we also develop a complementary approach based on product embedding to show global convergence. When applied to popular models such as 1-bit compressed sensing with Gaussian A and zero τ and the dithered 1-bit/multi-bit models with sub-Gaussian A and uniform dither τ, our unified treatment yields error rates that improve on or match the sharpest results in all instances. Particularly, PGD achieves the information-theoretic optimal rate O˜(kmL) for recovering k-sparse signals, and the rate O˜((kmL)1/3) for effectively sparse signals. For 1-bit compressed sensing of sparse signals, our result recovers the optimality of normalized binary iterative hard thresholding (NBIHT) that was proved very recently.
Asking appropriate follow-up questions is an important aspect of semi-structured interviews, yet interviewers’ knowledge and skills may limit their depth and affect outcomes. While many studies have demonstrated the usefulness of large language models (LLMs) for qualitative analysis, their potential in data collection remains underexplored. We adopt an AI-driven/Reverse “Wizard-of-Oz” setup to investigate how real-time LLM support for generating follow-up questions shapes semi-structured interviews. Through a study with 17 participants, we examine the value of LLM-generated follow-up questions, the evolving division of roles, relationships, collaborative behaviors, and responsibilities between interviewers and AI. Our findings (1) provide empirical evidence of the strengths and limitations of AI-generated follow-up questions (AGQs); (2) introduce a Human-AI collaboration framework in this interview context; and (3) propose human-centered design guidelines for AI-assisted interviews. We position LLMs as complements, not replacements, human judgment, and highlight pathways for integrating AI into qualitative data collection.
We review recent progress in understanding certain aspects of condensed matter systems from a high energy theory perspective. We discuss effective field theories that describe collective bulk and localized excitations in a variety of solid and fluid systems. Particular emphasis is placed on the role played by spacetime symmetries and their spontaneous breaking. The resulting Goldstone dynamics can be seen as underlying a wide variety of phenomena. We attempt to bridge the language gap between subfields while underscoring the numerous conceptual similarities.
Tax authorities often leave the consequences of non-compliance unspecified. We examine how disclosing those consequences affects tax payment and income reporting in four experiments. In a field experiment with HM Revenue and Customs, letters randomized to more than 200,000 taxpayers with outstanding income-tax debts increased payment when they warned of unspecified further action. Messages specifying financial or administrative costs did not produce the same benefit. An online payment experiment with 2050 taxpayers across fourteen conditions, including 1194 in the principal factorial design, found that revealing the maximum fine lowered the stated likelihood of paying. The reduction occurred at both modest and severe revealed maxima. It also occurred among participants who learned that the fine exceeded their prior estimate, contrary to the prediction from belief updating. An incentivized experiment with 827 participants extended the effect to income declarations with money at stake, although declarations differed between the two disclosed maxima in this setting. A laboratory experiment with 239 participants recorded differences in pupil dilation across enforcement messages without detectable differences in measured visual attention. Together, the findings suggest that ambiguity about enforcement consequences contributes to compliance. The physiological evidence is consistent with an affective account of this relationship. Disclosing a maximum penalty can therefore have a behavioral cost that tax authorities should consider when communicating enforcement policy.
Acoustic features of child directed speech (CDS) in caregiver-child conversational interactions are known to capture young children’s attention to language input and facilitate language learning. Emerging evidence suggests that caregivers’ use of CDS-like features when reading to their child, known as caregiver oral reading prosody, plays an important role in supporting preschoolers’ storybook comprehension. Do associations between caregiver oral reading prosody and preschoolers’ language extend beyond real-time listening comprehension to broader language skills? Forty-one caregivers read a children’s book to their child (ages: 4–5 years) as they normally would at home. The caregivers’ oral reading prosody (i.e., intonation and timing features) was examined in relation to children’s receptive vocabulary and core language skills. Moderated mediation models then investigated possible impacts of caregiver reading skills and shared reading time on these associations. Caregiver intonation range positively related to child vocabulary and caregiver inappropriate (i.e., ungrammatical) pausing negatively related to child core language skills. Mediation models revealed that caregiver reading skills indirectly related to child language through oral reading prosody measures. Notably, the indirect effect linking caregiver reading skills to child vocabulary via caregiver intonation range was moderated by shared reading time, with the strongest effects among dyads reporting less shared reading engagement. Findings illuminate caregiver oral reading prosody as one factor linking caregiver reading skills to child language, with particular relevance among families of caregivers with reading difficulties and for those with limited shared reading opportunities.