Prompting sometimes invokes expert-like downward shifts in multimodal models’ conceptual hierarchies
Bibliographic record
Abstract
Humans tend to privilege an intermediate level of categorization, known as the basic level, when categorizing objects that exist in a conceptual hierarchy (e.g. choosing to call a Labrador a dog instead of Labrador or animal). Domain experts demonstrate a downward shift in their object categorization behaviour, recruiting subordinate levels in a conceptual hierarchy as readily as conventionally basic categories (Tanaka & Philibert, 2022; Tanaka & Taylor, 1991). Do multimodal large language models show similar behavioural changes when prompted to behave in an expert-like way? We test whether GPT-4 with Vision (GPT-4V, OpenAI, 2023a) and LLaVA-1.5 (Liu, Li, Wu, & Lee, 2023; Liu, Li, Li, & Lee, 2023) demonstrate downward shifts using an object naming task and eliciting expert-like personas by altering the model’s system prompt. We find evidence of downward shifts in GPT-4V when expert system prompts are used, suggesting that human expert-like behaviour can be elicited from GPT-4V using prompting, but find no evidence of downward shift in LLaVA. We also find that there is an unpredicted upward shift in areas of non-expertise in some cases. These findings suggest that in the default case, GPT-4V is not a novice: instead, it behaves at default with a median level of expertise, while further expertise can be primed or forgotten through textual prompts. These results open the door for GPT-4V and similar models to be used as tools for studying differences in the behaviour of experts and novices, and even comparing contrasting levels of expertise within the same large language model.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.000 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.001 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.000 |
| Open science | 0.002 | 0.006 |
| Research integrity | 0.000 | 0.001 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".