CogFunSearch
FunSearch adapted to write short, readable programs that model how animals choose.
The loop
- Proposes
Python programs that predict an animal's next choice from its past choices and rewards
- Judged by
Scoring each program on how well it predicts recorded rat choices
- Steers
The best-scoring programs are fed back into the prompt, as in FunSearch
Headline result
In a bioRxiv preprint, on rat choices in a 2-armed drifting bandit task (data from Miller et al. 2021), CogFunSearch reliably found programs that beat the best existing cognitive model the authors compared against, both when improving supplied programs and when starting from scratch.
What this does not prove
A simple unconstrained neural network predicted significantly better than both the classic models and the discovered programs, so interpretability is gained at some cost in accuracy. Preprint, one task family.
How it generalizes the harness
The evolutionary loop finds a model of behavior, and the discovered program is a readable scientific artifact usable without the LLM.
Classification
| What it proposes | Program |
|---|---|
| Generative model | Pretrained LLM, frozen (inherited from FunSearch) |
| Algorithmic authority | Held-out predictive score of each program on behavioral data |
| Feedback | Score |
| What gets updated | Prompt context and population of candidates |
| When it runs | At inference time |
| Validity | Empirical (measured) |
| Evidence | Experiments reported |
| Lineage | Application |