Continuous glucose monitors (CGMs) used in diabetes care collect rich personal health data that could improve day-to-day self-management. However, current patient platforms only offer static summaries which do not support inquisitive user queries. Large language models (LLMs) could enable free-form inquiries about continuous glucose data, but deploying them over sensitive health records raises privacy and accuracy concerns. In this paper, we present **CGM-Agent**, a privacy-preserving framework for question answering over personal glucose data. In our design, the LLM serves purely as a reasoning engine that selects analytical functions. All computation occurs locally, and personal health data never leaves the user’s device. For evaluation, we construct a benchmark of 4,180 questions combining parameterized question templates with real user queries and ground truth derived from deterministic program execution. Evaluating 6 leading LLMs, we find that top models achieve 94% value accuracy on synthetic queries and 88% on ambiguous real-world queries. Errors stem primarily from intent and temporal ambiguity rather than computational failures. Additionally, lightweight models achieve competitive performance in our agent design, suggesting opportunities for low-cost deployment. We release our code and benchmark to support future work on trustworthy health agents.

In this work, we propose a Multi-LLM summarization framework, and investigate two different multi-LLM strategies including centralized and decentralized. Our multi-LLM summarization framework has two fundamentally important steps at each round of conversation: generation and evaluation. These steps are different depending on whether our multi-LLM decentralized summarization is used or centralized. In both our multi-LLM decentralized and centralized strategies, we have k different LLMs that generate diverse summaries of the text. However, during evaluation, our multi-LLM centralized summarization approach leverages a single LLM to evaluate the summaries and select the best one whereas k LLMs are used for decentralized multi-LLM summarization. Overall, we find that our multi-LLM summarization approaches significantly outperform the baselines that leverage only a single LLM by up to 3x. These results indicate the effectiveness of multi-LLM approaches for summarization.

While developing a story, novices and published writers alike have had to look outside themselves for inspiration. Language models have recently been able to generate text fluently, producing new stochastic narratives upon request. However, effectively integrating such capabilities with human cognitive faculties and creative processes remains challenging. We propose to investigate this integration with a multimodal writing support interface that offers writing suggestions textually, visually, and aurally. We conduct an extensive study that combines elicitation of prior expectations before writing, observation and semi-structured interviews during writing, and outcome evaluations after writing. Our results illustrate individual and situational variation in machine-in-the-loop writing approaches, suggestion acceptance, and ways the system is helpful. Centrally, we report how participants perform integrative leaps, by which they do cognitive work to integrate suggestions of varying semantic relevance into their developing stories. We interpret these findings, offering modeling and design recommendations for future creative writing support technologies.

This paper presents the system description by team niksss for the Qur’an QA 2022 Shared Task. The goal of this shared task was to evaluate systems for Arabic Reading Comprehension over the Holy Quran. The task was set up as a question-answering task, such that, given a passage from the Holy Quran (consisting of consecutive verses in a specific surah(Chapter)) and a question (posed in Modern Standard Arabic (MSA)) over that passage, the system is required to extract a span of text from that passage as an answer to the question. The span was required to be an exact sub-string of the passage. We attempted to solve this task using three techniques namely conditional text-to-text generation, embedding clustering, and transformers-based question answering.

This paper presents the 10th and 11th place system for Subtask A -English and Subtask A Arabic respectively of the SemEval 2022 -Task 6. The purpose of the Subtask A was to classify a given text sequence into sarcastic and nonsarcastic. We also breifly cover our method for Subtask B which performed subpar when compared with most of the submissions on the official leaderboard . All of the developed solutions used a transformers based language model for encoding the text sequences with necessary changes of the pretrained weights and classifier according to the language and subtask at hand .

This paper describes the system description for the HinglishEval challenge at INLG 2022. The goal of this task was to investigate the factors influencing the quality of the code- mixed text generation system. The task was divided into two subtasks, quality rating prediction and annotators’ disagreement prediction of the synthetic Hinglish dataset. We attempted to solve these tasks using sentence-level embeddings, which are obtained from mean pooling the contextualized word embeddings for all input tokens in our text. We experimented with various classifiers on top of the embeddings produced for respective tasks. Our best-performing system ranked 1st on subtask B and 3rd on subtask A. We make our code available here: https://github.com/nikhilbyte/Hinglish-qEval

Sanskrit is a classical language with about 30 million extant manuscripts fit for digitisation, available in written, printed or scanned-image forms. However, it is still considered to be a low-resource language when it comes to available digital resources. In this work, we release a post-OCR text correction dataset containing around 218,000 sentences, with 1.5 million words, from 30 different books. Texts in Sanskrit are known to be diverse in terms of their linguistic and stylistic usage since Sanskrit was the ‘lingua francua’ for discourse in the Indian subcontinent for about 3 millennia. Keeping this in mind, we release a multi-domain dataset, from areas as diverse as astronomy, medicine and mathematics, with some of them as old as 18 centuries. Further, we release multiple strong baselines as benchmarks for the task, based on pre-trained Seq2Seq language models. We find that our best-performing model, consisting of byte level tokenization in conjunction with phonetic encoding (Byt5+SLP1), yields a 23% point increase over the OCR output in terms of word and character error rates. Moreover, we perform extensive experiments in evaluating these models on their performance and analyse common causes of mispredictions both at the graphemic and lexical levels. Our code and dataset is publicly available at https://github.com/ayushbits/pe-ocr-sanskrit.

This paper describes the 9th place system description for SemEval-2022 Task 7. The goal of this shared task was to develop computational models to predict how plausible a clarification made on an instructional text is. This shared task was divided into two Subtasks A and B. We attempted to solve these using various transformers-based architecture under different regime. We initially treated this as a text2text generation problem but comparing it with our recent approach we dropped it and treated this as a text-sequence classification and regression depending on the Subtask.