mlsanityData to 7 Oct 2026

Evidence of Learned Look-Ahead in a Chess-Playing Neural Network

arXiv 2406.00877 ↗First posted 2 Jun 2024cs.LG cs.AIcited by 33

Not on the map. Why?

Abstract (arXiv)

Do neural networks learn to implement algorithms such as look-ahead or search "in the wild"? Or do they rely purely on collections of simple heuristics? We present evidence of learned look-ahead in the policy network of Leela Chess Zero, the currently strongest neural chess engine. We find that Leela internally represents future optimal moves and that these representations are crucial for its final output in certain board states. Concretely, we exploit the fact that Leela is a transformer that treats every chessboard square like a token in language models, and give three lines of evidence (1) activations on certain squares of future moves are unusually important causally; (2) we find attention heads that move important information "forward and backward in time," e.g., from squares of future moves to squares of earlier ones; and (3) we train a simple probe that can predict the optimal move 2 turns ahead with 92% accuracy (in board states where Leela finds a single best line). These findings are an existence proof of learned look-ahead in neural networks and might be a step towards a better understanding of their capabilities.

Key facts. Cited by 33 papers in the map (data through 7 Oct 2026). Builds on 23 of its 47 references that are in the map. First posted on arXiv 2 Jun 2024.

Builds on

23 of 47 references.

References · 23

  1. 2017Attention Is All You Needcited by 59,932
  2. 2022Locating and Editing Factual Associations in GPTcited by 2,609
  3. 2022Transformers learn in-context by gradient descentcited by 714
  4. 2022What learning algorithm is in-context learning? Investigations with linear modelscited by 671
  5. 2019Designing and Interpreting Probes with Control Taskscited by 608
  6. 2023Eliciting Latent Predictions from Transformers with the Tuned Lenscited by 579
  7. 2016RL2: Fast Reinforcement Learning via Slow Reinforcement Learningcited by 490
  8. 2022Emergent World Representations: Exploring a Sequence Model Trained on a Synthetic Taskcited by 417
  9. 2021Causal Abstractions of Neural Networkscited by 401
  10. 2016Learning to reinforcement learncited by 369
  11. 2023Towards Best Practices of Activation Patching in Language Models: Metrics and Methodscited by 328
  12. 2023Emergent Linear Representations in World Models of Self-Supervised Sequence Modelscited by 313
11 more
  1. 2019Risks from Learned Optimization in Advanced Machine Learning Systemscited by 250
  2. 2023Supervised Pretraining Can Learn In-Context Reinforcement Learningcited by 137
  3. 2021Acquisition of Chess Knowledge in AlphaZerocited by 120
  4. 2023Future Lens: Anticipating Subsequent Tokens from a Single Hidden Statecited by 116
  5. 2023Jump to Conclusions: Short-Cutting Transformers With Linear Transformationscited by 89
  6. 2023ChessGPT: Bridging Policy Learning and Language Modelingcited by 73
  7. 2021Chess as a Testbed for Language Model State Trackingcited by 69
  8. 2024Emergent World Models and Latent Variable Estimation in Chess-Playing Language Modelscited by 60
  9. 2024A Mechanistic Analysis of a Transformer Trained on a Symbolic Multi-Step Reasoning Taskcited by 49
  10. 2023Bridging the Human-AI Knowledge Gap: Concept Discovery and Transfer in AlphaZerocited by 41
  11. 2020The Chess Transformer: Mastering Play using Generative Language Modelscited by 36

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2024: 8; 2025: 8; 2026: 17.

33 papers.

Sort:

By area:

  1. 2024ICLR: In-Context Learning of Representationscited by 49
  2. 2024Unlocking the Future: Exploring Look-Ahead Planning Mechanistic Interpretability in Large Language Modelscited by 21
  3. 2024Transformers Struggle to Learn to Searchcited by 19
  4. 2024Planning in a recurrent neural network that plays Sokobancited by 17
  5. 2025How Do LLMs Use Their Depth?cited by 15
  6. 2024Mastering Chess with a Transformer Modelcited by 14
  7. 2024Transformers Use Causal World Models in Maze-Solving Taskscited by 13
  8. 2026Mechanistic Interpretability for Large Language Model Alignment: Progress, Challenges, and Future Directionscited by 8
  9. 2024On Logical Extrapolation for Mazes with Recurrent and Implicit Networkscited by 7
  10. 2026Chessformer: A Unified Architecture for Chess Modelingcited by 6
  11. 2025Path Channels and Plan Extension Kernels: a Mechanistic Description of Planning in a Sokoban RNNcited by 4
  12. 2024Predicting Chess Puzzle Difficulty with Transformerscited by 3
21 more
  1. 2025Exploring Human-AI Conceptual Alignment through the Prism of Chesscited by 2
  2. 2025ParaScopes: What do Language Models Activations Encode About Future Text?cited by 2
  3. 2026Tracing the Thought of a Grandmaster-level Chess-Playing Transformercited by 2
  4. 2025K-MSHC: Unmasking Minimally Sufficient Head Circuits in Large Language Models with Experiments on Syntactic Classification Taskscited by 1
  5. 2025Detecting and Characterizing Planning in Language Modelscited by 1
  6. 2026Mixture of Masters: Sparse Chess Language Models with Player Routingcited by 1
  7. 2026Interaction Locality in Hierarchical Recursive Reasoningcited by 1
  8. 2026Planning as Emergent Behavior in Reinforcement Learning with Relational Hidden Statescited by 1
  9. 2025Understanding the learned look-ahead behavior of chess neural networkscited by 0
  10. 2025Liquid Reasoning Transformers: A Sudoku-Based Prototype for Chess-Scale Algorithmic Taskscited by 0
  11. 2026Abstract Concept Modelling in Conceptual Spaces: A Study on Chess Strategiescited by 0
  12. 2026Large Language Models Can Take False First Steps at Inference-time Planningcited by 0
  13. 2026Understanding the Emergence of Seemingly Useless Features in Next-Token Predictorscited by 0
  14. 2026Where's the Plan? Locating Latent Planning in Language Models with Lightweight Mechanistic Interventionscited by 0
  15. 2026A Close Look At World Model Recovery In Supervised Fine-Tuned LLM Plannerscited by 0
  16. 2026Latent Programming Horizons in Coding Agentscited by 0
  17. 2026Beyond Search-Imitation: Prior-Directed Exploration for Searchless Chesscited by 0
  18. 2026Exploring Collaboration between a language and a non-language agentcited by 0
  19. 2026What Counts as Strategic Reasoning? A Systematic Mapping of Chess Research on Humans, Engines, and Language Modelscited by 0
  20. 2026Increasing Skill Level Recruits Deeper Attention Layers in a Frozen Chess Transformercited by 0
  21. 2026Understanding Decision-Making Mechanisms in Neural Routing Solverscited by 0