Khizar Qureshi

Khizar Qureshi, Tauhid Zaman

We study the problem of predicting the future performance of cryptocurrencies using social media data. We propose a new model to measure the engagement of users with topics discussed on social media based on interactions with social media posts. This model overcomes the limitations of previous volume and sentiment based approaches. We use this model to estimate engagement coefficients for 48 cryptocurrencies created between 2019 and 2021 using data from Twitter from the first month of the cryptocurrencies' existence. We find that the future returns of the cryptocurrencies are dependent on the engagement coefficients. Cryptocurrencies whose engagement coefficients are too low or too high have lower returns. Low engagement coefficients signal a lack of interest, while high engagement coefficients signal artificial activity which is likely from automated accounts known as bots. We measure the amount of bot posts for the cryptocurrencies and find that generally, cryptocurrencies with more bot posts have lower future returns. While future returns are dependent on both the bot activity and engagement coefficient, the dependence is strongest for the engagement coefficient, especially for short-term returns. We show that simple investment strategies which select cryptocurrencies with engagement coefficients exceeding a fixed threshold perform well for holding times of a few months.

Khizar Qureshi, Geoffrey Martin, Yifan Peng

A key challenge for large language models is token cost per query and overall deployment cost. Clinical inputs are long, heterogeneous, and often redundant, while downstream tasks are short and high stakes. We study budgeted context selection, where a subset of document units is chosen under a strict token budget so an off-the-shelf generator can meet fixed cost and latency constraints. We cast this as a knapsack-constrained subset selection problem with two design choices, unitization that defines document segmentation and selection that determines which units are kept. We propose \textbf{RCD}, a monotone submodular objective that balances relevance, coverage, and diversity. We compare sentence, section, window, and cluster-based unitization, and introduce a routing heuristic that adapts to the budget regime. Experiments on MIMIC discharge notes, Cochrane abstracts, and L-Eval show that optimal strategies depend on the evaluation setting. Positional heuristics perform best at low budgets in extractive tasks, while diversity-aware methods such as MMR improve LLM generation. Selector choice matters more than unitization, with cluster-based grouping reducing performance and other schemes behaving similarly. ROUGE saturates for LLM summaries, while BERTScore better reflects quality differences. We release our code at https://github.com/stone-technologies/ACL_budget_paper.