Search for: All records

Vision-language models are integral to computer vision research, yet many high-performing models remain closed-source, obscuring their data, design and training recipe. The research community has responded by using distillation from black-box models to label training data, achieving strong benchmark results, at the cost of measurable scientific progress. However, without knowing the details of the teacher model and its data sources, scientific progress remains difficult to measure. In this paper, we study building a Perception Language Model (PLM) in a fully open and reproducible framework for transparent research in image and video understanding. We analyze standard training pipelines without distillation from proprietary models and explore large-scale synthetic data to identify critical data gaps, particularly in detailed video understanding. To bridge these gaps, we release 2.8M human-labeled instances of fine-grained video question-answer pairs and spatio-temporally grounded video captions. Additionally, we introduce PLM-VideoBench, a suite for evaluating challenging video understanding tasks focusing on the ability to reason about "what", "where", "when", and "how" of a video. We make our work fully reproducible by providing data, training recipes, code & models. 

In combination with contextualized human interpretation, computational text analysis offers a quantitative approach to interrogating the nature of participation and social positioning in discourse. Using meeting transcript data from the development of a co-design research-practice partnership, we examine the roles and forms of participation that contribute to an effective collaboration between a multileveled school system and researcher partners. We apply computational methods to explore the language of co-design and multi-stakeholder perspectives in support of educational improvement science efforts and our theoretical understanding of partnership roles. Results indicate participation patterns align with documented roles in codesign partnerships and highlight the space dedicated to process reflection, context sharing, and logistical coordination 

The learning sciences community is currently exploring new ways to enact productive and equitable co-design research-practice partnerships that are sensitive to all the concerns and needs of stakeholders. The paper contributes to that still-growing literature through an interaction analysis of a co-design discussion involving school district partners that unfolded about cultural relevance and sensitivity in relation to the use of a specific image in an elementary school coding lesson. The episode involved looking moment-by-moment at how district educators recognized and acknowledged that a specific design decision could be harmful for a minoritized population of students enrolled in the district. However, once a key change was made to be more culturally responsive and considerate, new and unexpected pedagogical challenges appeared. This case serves to illustrate some of the unexpected tensions that can appear in real-time when unanticipated questions about cultural relevance are foregrounded during lesson and materials co-design. 

The learning sciences community is currently exploring new ways to enact productive and equitable co-design research-practice partnerships that are sensitive to all the concerns and needs of stakeholders. The paper contributes to that still-growing literature through an interaction analysis of a co-design discussion involving school district partners that unfolded about cultural relevance and sensitivity in relation to the use of a specific image in an elementary school coding lesson. The episode involved looking moment-by-moment at how district educators recognized and acknowledged that a specific design decision could be harmful for a minoritized population of students enrolled in the district. However, once a key change was made to be more culturally responsive and considerate, new and unexpected pedagogical challenges appeared. This case serves to illustrate some of the unexpected tensions that can appear in real-time when unanticipated questions about cultural relevance are foregrounded during lesson and materials co-design. 

In this chapter, we share observations from a multiyear design-based research project exploring how to teach developmentally appropriate coding concepts and skills in kindergarten. We focus on coding toys that fit within a genre we call “grid- agent” robot coding toys. These are robots that are specifically for early childhood, commercially available, screen-free, tangible, moveable and programmable. Grid- agent robot toys invite children to explore mathematics through precise movements across a grid space. 

Abstract The use of transparent test/source masses can benefit future measurements of Newton’s gravitational constant G . Such transparent test mass materials can enable nondestructive, quantitative internal density gradient measurements using optical interferometry and allow in-situ optical metrology methods to be realized for the critical distance measurements often needed in a G apparatus. To confirm the sensitivity of such optical interferometry measurements to internal density gradients it is desirable to conduct a check with a totally independent technique. We present an upper bound on possible internal density gradients in lead tungstate (PbWO 4 ) crystals using a Talbot-Lau neutron interferometer on the Cold Neutron Imaging Facility at NIST. We placed an upper bound on a fractional atomic density gradient in two PbWO 4 test crystals of 1 N d N d x < 0.5 × 10 − 6  cm −1 . This value is about two orders of magnitude smaller than required for G measurements. We discuss the implications of this result and of other nondestructive methods for characterization of internal density inhomogeneties which can be applied to test masses in G experiments. 

Abstract: This paper reports on a study of the dynamics of a Research-Practice Partnership (RPP) oriented around design, specifically the co-design model. The RPP is focused on supporting elementary school computer science (CS) instruction by involving paraprofessional educators and teachers in curricular co-design. A problem of practice addressed is that few elementary educators have backgrounds in teaching CS and have limited available instructional time and budget for CS. The co-design strategy entailed highlighting CS concepts in the mathematics curriculum during classroom instruction and designing computer lab lessons that explored related ideas through programming. Analyses focused on tensions within RPP interaction dynamics and how they were accommodated when RPP partners were designing for co-design activities, a critical component that leads to curricular co-design itself. We illustrate these tensions with examples of clusters of activity that appeared repeatedly among the research and practice team members when “designing for co-design”. 

The authors show how a toy robot was used with a number line in a Year 1 class to explore equality and missing addends. They conclude that coding toys can be used to integrate mathematics and programming in engaging and innovative ways, to support young children in their learning. 

The purpose of this study was to explore how kindergarten students (aged 5–6 years) engaged with mathematics as they learned programming with robot coding toys. We video-recorded 16 teaching sessions of kindergarten students’ (N = 36) mathematical and programming activities. Students worked in small groups (4–5 students) with robot coding toys on the floor in their classrooms, solving tasks that involved programming these toys to move to various locations on a grid. Drawing on a semiotic mediation perspective, we analyzed video data to identify the mathematics concepts and skills students demonstrated and the overlapping mathematics-programming knowledge exhibited by the students during these programming tasks. We found that kindergarten children used spatial, measurement, and number knowledge, and the design of the tasks, affordances of the robots, and types of programming knowledge influenced how the students engaged with mathematics. The paper concludes with a discussion about the intersections of mathematics and programming knowledge in early childhood, and how programming robot toys elicited opportunities for students to engage with mathematics in dynamic and interconnected ways, thus creating an entry point to reassert mathematics beyond the traditional school content and curriculum.