Existing toolkits for data physicalization prioritize ease of use and adoption by novices. This is often achieved by limiting the affordances of the materials used and constraining design possibilities. The result is limited opportunities for creating expressive physicalizations. To address this limitation and to better understand how to support the creation process of expressive physicalizations, we created ClayPhys, a low-fidelity data physicalization toolkit designed to encourage making expressive data physicalizations. Our toolkit consists of clay, clay work tools, instruction and documentation handbooks, and warm-up activities that scaffold the design process. We studied the use of ClayPhys in a one-day workshop with nine expert participants. From our analysis of participants’ created data physicalizations, we observed that using ClayPhys, participants could map data to different visual and physical variables, and that their designs incorporated various data interaction styles. Informed by our findings, we discuss implications for designing higher-fidelity expressive data physicalization toolkits.

Blind and low-vision (BLV) employees in mixed-visual ability teams often encounter information (e.g., PDFs, diagrams) in inaccessible formats. To enable teamwork, teams must transform these representations by modifying or re-creating them into accessible forms. However, these transformations are frequently overlooked, lack infrastructural support, and cause additional labour. To design systems that move beyond one-off accommodations to effective mixed-ability collaboration, we need a deeper understanding of the representations, their transformations and how they occur. We conducted a week-long diary study with follow-up interviews with 23 BLV and sighted professionals from five legal, non-profit, and consulting teams, documenting 36 transformation cases. Our analysis characterizes how teams perform representational transformations for accessibility: how they are triggered proactively or reactively, how they simplify or enhance, and four common patterns in which workers coordinate with each other to address representational incompatibility. Our findings uncover opportunities for designing systems that can better support mixed-visual ability work.

Sonification serves as a powerful tool for data accessibility, especially for people with vision loss. Among various modalities, speech is a familiar means of communication similar to the role of text in visualization. However, speech-based sonification is underexplored. We introduce SpeechTone, a novel speech-based mark for data sonification and extension to the existing Erie declarative grammar for sonification. It encodes data into speech attributes such as pitch, speed, voice and speech content. We demonstrate the efficacy of SpeechTone through three examples.

Diagrams often appear as node-link representations in contexts such as taxonomies, mind maps and networks in textbooks. Despite their pervasiveness, they present accessibility challenges for blind and low-vision people. To address this challenge, we introduce Touch-and-Audio-based Diagram Access (TADA), a tablet-based interactive system that makes diagram exploration accessible through musical tones and speech. We designed TADA informed by an interview study with 15 participants who shared their challenges and strategies with diagrams. TADA enables people to access a diagram by: i) engaging in open-ended touch-based explorations, ii) searching for nodes, iii) navigating between nodes and iv) filtering information. We evaluated TADA with 25 participants and found it useful for gaining different perspectives on diagrammatic information.

Step-by-step reasoning methods, such as the Chain-of-Thought (CoT), have been demonstrated to be highly effective in harnessing the reasoning capabilities of Large Language Models (LLMs). Recent research efforts have sought to distill LLMs into Small Language Models (SLMs), with a significant focus on transferring the reasoning capabilities of LLMs to SLMs via CoT. However, the outcomes of CoT distillation are inadequate for knowledge-intensive reasoning tasks. This is because generating accurate rationales requires crucial factual knowledge, which SLMs struggle to retain due to their parameter constraints. We propose a retrieval-based CoT distillation framework, named Probe then Retrieve and Reason (PRR), which distills the question probing and reasoning capabilities from LLMs into SLMs. We train two complementary distilled SLMs, a probing model and a reasoning model, in tandem. When presented with a new question, the probing model first identifies the necessary knowledge to answer it, generating queries for retrieval. Subsequently, the reasoning model uses the retrieved knowledge to construct a step-by-step rationale for the answer. In knowledge-intensive reasoning tasks, such as StrategyQA and OpenbookQA, our distillation framework yields superior performance for SLMs compared to conventional methods, including simple CoT distillation and knowledge-augmented distillation using raw questions.

Aspect Sentiment Triplet Extraction (ASTE) aims at extracting triplets from a given sentence, where each triplet includes an aspect, its sentiment polarity, and a corresponding opinion explaining the polarity. Existing methods are poor at detecting complicated relations between aspects and opinions as well as classifying multiple sentiment polarities in a sentence. Detecting unclear boundaries of multi-word aspects and opinions is also a challenge. In this paper, we propose a Multi-Task Dual-Tree Network (MTDTN) to address these issues. We employ a constituency tree and a modified dependency tree in two sub-tasks of Aspect Opinion Co-Extraction (AOCE) and ASTE, respectively. To enhance the information interaction between the two sub-tasks, we further design a Transition-Based Inference Strategy (TBIS) that transfers the boundary information from tags of AOCE to ASTE through a transition matrix. Extensive experiments are conducted on four popular datasets, and the results show the effectiveness of our model.

Recently, prompt tuning has shown remarkable performance as a new learning paradigm, which freezes pre-trained language models (PLMs) and only tunes some soft prompts. A fixed PLM only needs to be loaded with different prompts to adapt different downstream tasks. However, the prompts associated with PLMs may be added with some malicious behaviors, such as backdoors. The victim model will be implanted with a backdoor by using the poisoned prompt. In this paper, we propose to obtain the poisoned prompt for PLMs and corresponding downstream tasks by prompt tuning. We name this Poisoned Prompt Tuning method "PPT". The poisoned prompt can lead a shortcut between the specific trigger word and the target label word to be created for the PLM. So the attacker can simply manipulate the prediction of the entire model by just a small prompt. Our experiments on various text classification tasks show that PPT can achieve a 99% attack success rate with almost no accuracy sacrificed on original task. We hope this work can raise the awareness of the possible security threats hidden in the prompt.