Unverified50% confidenceFactExact time
A 2023 MIT study trained a GPT model to predict the next move in Othello games and, through probing techniques, found that the model's internal representations encoded the complete board state.
1
Sources
50%
Confidence
Long-term
Relevance
9/6/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedOpenAI's GPT series is based on an autoregressive language model architecture, with the core training paradigm being next-token probability prediction.71% similarUnverifiedOthello GPT研究表明，仅学习黑白棋合法走法的模型在内部追踪了整个棋盘状态，可用探针读出71% similarUnverifiedThe GPT series of models uses an autoregressive generation approach, predicting the next most likely output token by token.70% similarVerifiedThe GPT series adopts the autoregressive training objective of "predicting the next word".70% similarUnverifiedThe self-training breakthrough in GPT-5.3 extended the self-evaluation mechanism to the entire training data pipeline, meaning the training materials themselves are also generated by the model, not just used for evaluation.70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/865874API
curl https://kongchang.com/api/v1/knowledge/claims/865874MCP
get_claim(id=865874)