Adapting BERT for 'Apples to Apples' Gameplay
Bibliographic record
Abstract
"Apples to Apples" is a word association game where players match red noun cards with a given green adjective card in the hopes of winning the round through the selection of their card by the round judge. There are many possible selection criteria for determining which red card to play in a round. Combinations may focus solely on the direct association between a red and green card, on the humour elicited by a card pair, or through a wide variety of other selection criteria. The game's difficulty comes from the anticipation and matching of the card played by a player to the criterion of the current round judge. Our aim was to create an agent capable of playing "Apples to Apples," able to discover the current judge criterion and play the card in hand that best matches that criterion. To this end, we decided to explore the use of BERT pretrained models and evaluate their viability in word game agents. In this study, we leveraged fine tuned BERT models and developed a Naive Bayes classifier to simulate and predict judge personalities in "Apples to Apples." Beyond these main contributions, this work also involved significant efforts in creating supporting tools for both training and testing processes. Specifically, we manually annotated nearly 1000 card pairs using a custom annotation program to reduce human error and fatigue and developed an environment testing tool to ensure platform consistency during training and testing. These contributions highlight our focus on both data quality and system robustness.
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 machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.002 | 0.008 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.000 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.002 | 0.002 |
| Open science | 0.002 | 0.001 |
| Research integrity | 0.001 | 0.002 |
| Insufficient payload (model declined to judge) | 0.003 | 0.002 |
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 source (direct Gemma or distilled Codex), 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".