REVIEW 5 cited by
Do Large Language Models Perform the Way People Expect? Measuring the Human Generalization Function
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
Signed reviews
read the original abstract
What makes large language models (LLMs) impressive is also what makes them hard to evaluate: their diversity of uses. To evaluate these models, we must understand the purposes they will be used for. We consider a setting where these deployment decisions are made by people, and in particular, people's beliefs about where an LLM will perform well. We model such beliefs as the consequence of a human generalization function: having seen what an LLM gets right or wrong, people generalize to where else it might succeed. We collect a dataset of 19K examples of how humans make generalizations across 79 tasks from the MMLU and BIG-Bench benchmarks. We show that the human generalization function can be predicted using NLP methods: people have consistent structured ways to generalize. We then evaluate LLM alignment with the human generalization function. Our results show that -- especially for cases where the cost of mistakes is high -- more capable models (e.g. GPT-4) can do worse on the instances people choose to use them for, exactly because they are not aligned with the human generalization function.
Forward citations
Cited by 5 Pith papers
-
Potemkin Understanding in Large Language Models
LLMs frequently pass definition questions yet fail to use the same concepts in classification, generation, and editing tasks, a gap the authors call potemkin understanding.
-
Evolution and The Knightian Blindspot of Machine Learning
ML's formalisms, particularly RL's, exclude Knightian uncertainty, and evolution's diversify-and-filter mechanisms point toward a direct remedy.
-
DSADF: Thinking Fast and Slow for Decision Making
A dual-system framework where an RL agent handles familiar subtasks and a VLM planner/executor handles unfamiliar ones improves task success and speed in Crafter and Housekeep.
-
HeartbeatCam: Self-Triggered Photo Elicitation of Stress Events Using Wearable Sensing
A smartwatch-triggered AR-glasses capture system records sparse image-audio clips during elevated stress for later therapy review.
-
Revisiting Rogers' Paradox in the Context of Human-AI Interaction
A simulation of Rogers' Paradox with an AI agent that learns the population average shows that cheap AI alone does not improve collective world understanding, while critical appraisal and independent AI learning can.
Discussion (0). Continue with ORCID to comment.