Arrow Research search

Author name cluster

Paul Glasserman

Possible papers associated with this exact author name in Arrow. This page groups case-insensitive exact name matches and is not a full identity disambiguation profile.

2 papers
1 author row

Possible papers

2

AAAI Conference 2026 Short Paper

Do Large Language Models (LLMs) Understand Chronology? (Student Abstract)

  • Pattaraphon Kenny Wongchamcharoen
  • Paul Glasserman

Large language models have shown great potential as forecasting tools in finance and economics, but backtesting performance is subject to look-ahead bias if the period overlaps with an LLM’s training window. Prompt-based attempts to avoid look-ahead bias require that LLMs understand chronology. We test LLMs’ ability to understand and enforce chronological order in three types of tasks: sorting randomly shuffled historical events; conditional sorting of events defined by some conditions; and anachronism detection based on intersections of multiple timelines. Our experiments use events that we first confirm are known to the LLM; this ensures that we test chronological understanding on an LLM’s pretrained internal knowledge. Across three LLM families— GPT-4.1 (standard), GPT-5 (hybrid-reasoning), and Claude 3.7 Sonnet (large-reasoning, with and without Extended Thinking), we find that performance degrades rapidly with problem complexity but improves greatly for reasoning models with test-time extended reasoning. These patterns are important for the real-time application of LLMs in finance.

TMLR Journal 2026 Journal Article

Privacy Profiles Under Tradeoff Composition

  • Paul Glasserman

Privacy profiles and tradeoff functions are two frameworks for comparing differential privacy guarantees of alternative privacy mechanisms. We study connections between these frameworks. We show that the composition of tradeoff functions corresponds to a binary operation on privacy profiles we call their T-convolution. Composition of tradeoff functions characterizes group privacy guarantees, so the T-convolution provides a bridge for translating group privacy properties from one framework to the other. Composition of tradeoff functions has also been used to characterize mechanisms with log-concave additive noise; we derive a corresponding property based on privacy profiles. We also derive new bounds on privacy profiles for log-concave mechanisms based on new convexity properties. In developing these ideas, we characterize regular privacy profiles, which are privacy profiles for mutually absolutely continuous probability measures.