ChessMoveLLM: Large Language Models for Chess Next Move Prediction
Bibliographic record
Abstract
Creating a chess engine using language models is challenging due to the complexity of understanding the logic and depth of moves. To solve this problem, we propose several approaches with different types of training of different LLMs that already have some knowledge of tasks and not necessarily on chess, and adjust them for faster learning. We propose three groups of methods and five approaches to enhance state-of-the-art results. Our approaches will involve different types of database manipulation, the training of different types of networks and the application of a new, widely used approach, the RAG (Retrieval-Augmented Generation). We demonstrated the potential of networks and approaches with promising results for the first 2 methods, and we demonstrated with RAG in combination with FAISS (Facebook AI Similarity Search) [1], algebraic notation and GPT3.5-turbo-instruct that it is possible to increase the ability of networks to predict good chess moves. The first approach demonstrated that successive move generation benefits generalization and reduces illegal moves; for the second approach, we proved that the use of noiseless algebraic notation gives the network the possibility of progressing significantly before giving illegal moves. The second group of approaches, demonstrated that using the FEN (Forsyth-Edwards Notation), a textual graphical representation, for the prediction of move sequences to force context understanding could give promising results with a low number of epochs. Even with limitations in computation time and epochs, we were able to find promising results with a part of our dataset that can be improved with more epochs and data.
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 | 0.000 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| 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".