Author name cluster

Todd Gureckis

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.

3 papers

1 author row

NeurIPS Conference 2025 Conference Paper

Do different prompting methods yield a common task representation in language models?

Guy Davidson
Todd Gureckis
Brenden Lake
Adina Williams

Demonstrations and instructions are two primary approaches for prompting language models to perform in-context learning (ICL) tasks. Do identical tasks elicited in different ways result in similar representations of the task? An improved understanding of task representation mechanisms would offer interpretability insights and may aid in steering models. We study this through function vectors (FVs), recently proposed as a mechanism to extract few-shot ICL task representations. We generalize FVs to alternative task presentations, focusing on short textual instruction prompts, and successfully extract instruction function vectors that promote zero-shot task accuracy. We find evidence that demonstration- and instruction-based function vectors leverage different model components, and offer several controls to dissociate their contributions to task performance. Our results suggest that different task prompting forms do not induce a common task representation through FVs but elicit different, partly overlapping mechanisms. Our findings offer principled support to the practice of combining instructions and task demonstrations, imply challenges in universally monitoring task inference across presentation forms, and encourage further examinations of LLM task inference mechanisms.

PDF Details

NeurIPS Conference 2017 Conference Paper

Question Asking as Program Generation

Anselm Rothe
Brenden Lake
Todd Gureckis

A hallmark of human intelligence is the ability to ask rich, creative, and revealing questions. Here we introduce a cognitive model capable of constructing human-like questions. Our approach treats questions as formal programs that, when executed on the state of the world, output an answer. The model specifies a probability distribution over a complex, compositional space of programs, favoring concise programs that help the agent learn in the current context. We evaluate our approach by modeling the types of open-ended questions generated by humans who were attempting to learn about an ambiguous situation in a game. We find that our model predicts what questions people will ask, and can creatively produce novel questions that were not present in the training set. In addition, we compare a number of model variants, finding that both question informativeness and complexity are important for producing human-like questions.

PDF Details

RLDM Conference 2013 Conference Abstract

Modeling active learning decisions during causal learning

Anna Coenen
Todd Gureckis
Bob Rehder

An important type of decision making concerns how people choose to gather information which reduces their uncertainty about the world. For example, when learning about a novel piece of technology, like a smartphone, people often actively intervene on various aspects in order to better understand the function of the system. Interventions allow us to tell apart causal structures that are indistinguishable through obser- vation, but only if the right variables are intervened on. Normative models of decision making developed in the machine learning literature specify a process of comparing hypotheses to identify those interventions that will allow a learner to distinguish between them. An experiment that asked subjects to decide between two causal hypotheses found that while they often chose useful interventions, they frequently perform inter- ventions whose expected effects were typical of one causal structure but that did not always allow the two structures to be distinguished. We interpret this tendency as a type of positive-test-strategy with a preference for outcomes that are representative of a single causal structure.

PDF Details