We introduce LLM CHESS, an evaluation framework designed to probe the generalization of reasoning and instruction-following abilities in large language models (LLMs) through extended agentic interaction in the domain of chess. We rank over 50 open and closed source models by playing against a random opponent using a range of behavioral metrics, including win and loss rates, move quality, move legality, hallucinated actions, and game duration. For a subset of top reasoning models, we derive an Elo estimate by playing against a chess engine with variably configured skill, which allows for comparisons between models in an easily understandable way. Despite the simplicity of the instruction-following task and the weakness of the opponent, many state-of-the-art models struggle to complete games or achieve consistent wins. Similar to other benchmarks on complex reasoning tasks, our experiments reveal a clear separation between reasoning and non-reasoning models. However, unlike existing static benchmarks, the stochastic and dynamic nature of LLM CHESS uniquely reduces overfitting and memorization while preventing benchmark saturation, proving difficult even for top reasoning models. To support future work on evaluating reasoning and instruction-following in LLMs, we release our experimental framework, a public leaderboard, and a dataset of associated games.

Language models are traditionally designed around causal masking. In domains with spatial or relational structure, causal masking is often viewed as inappropriate, and sequential linearizations are instead used. Yet the question of whether it is viable to accept the information loss introduced by causal masking on nonsequential data has received little direct study, in part because few domains offer both spatial and sequential representations of the same dataset. In this work, we investigate this issue in the domain of chess, which naturally supports both representations. We train language models with bidirectional and causal self-attention mechanisms on both spatial (board-based) and sequential (move-based) data. Our results show that models trained on spatial board states - \textit{even with causal masking} - consistently achieve stronger playing strength than models trained on sequential data. While our experiments are conducted on chess, our results are methodological and may have broader implications: applying causal masking to spatial data is a viable procedure for training unimodal LLMs on spatial data, and in some domains is even preferable to sequentialization.

Contemporary chess engines offer precise yet opaque evaluations, typically expressed as centipawn scores. While effective for decision-making, these outputs obscure the underlying contributions of individual pieces or patterns. In this paper, we explore adapting SHAP (SHapley Additive exPlanations) to the domain of chess analysis, aiming to attribute a chess engines evaluation to specific pieces on the board. By treating pieces as features and systematically ablating them, we compute additive, per-piece contributions that explain the engines output in a locally faithful and human-interpretable manner. This method draws inspiration from classical chess pedagogy, where players assess positions by mentally removing pieces, and grounds it in modern explainable AI techniques. Our approach opens new possibilities for visualization, human training, and engine comparison. We release accompanying code and data to foster future research in interpretable chess AI.

Your company's CEO is retiring. You search for a successor. You can promote an employee from the company familiar with the company's operations, or recruit an external professional manager. Who should you prefer? It has not been clear how to address this question, the "subject matter expertise vs. professional manager debate", quantitatively and objectively. We note that a company's success depends on long sequences of interdependent decisions, with often-opposing recommendations of diverse board members. To model this task in a controlled environment, we utilize chess - a complex, sequential game with interdependent decisions which allows for quantitative analysis of performance and expertise (since the states, actions and game outcomes are well-defined). The availability of chess engines differing in style and expertise, allows scalable experimentation. We considered a team of (computer) chess players. At each turn, team members recommend a move and a manager chooses a recommendation. We compared the performance of two manager types. For manager as "subject matter expert", we used another (computer) chess player that assesses the recommendations of the team members based on its own chess expertise. We examined the performance of such managers at different strength levels. To model a "professional manager", we used Reinforcement Learning (RL) to train a network that identifies the board positions in which different team members have relative advantage, without any pretraining in chess. We further examined this network to see if any chess knowledge is acquired implicitly. We found that subject matter expertise beyond a minimal threshold does not significantly contribute to team synergy. Moreover, performance of a RL-trained "professional" manager significantly exceeds that of even the best "expert" managers, while acquiring only limited understanding of chess.

There exist several techniques for representing the chess board inside the computer. In the first part of this paper, the concepts of the bitboard-representation and the advantages of (rotated) bitboards in move generation are explained. In order to illustrate those ideas practice, the concrete implementation of the move-generator in FUSc# is discussed and we explain a technique how to verify the move-generator with the "perft"-command. We show that the move-generator of FUSc# works 100% correct. The second part of this paper deals with reinforcement learning in computer chess (and beyond). We exemplify the progress that has been made in this field in the last 15-20 years by comparing the "state of the art" from 2002-2008, when FUSc# was developed, with recent innovations connected to "AlphaZero". We discuss how a "FUSc#-Zero" could be implemented and what would be necessary to reduce the number of training games necessary to achieve a good performance. This can be seen as a test case to the general prblem of improving "sample effciency" in reinforcement learning. In the final part, we move beyond computer chess, as the importance of sample effciency extends far beyond board games into a wide range of applications where data is costly, diffcult to obtain, or time consuming to generate. We review some application of the ideas developed in AlphaZero in other domains, i.e. the "other Alphas" like AlphaFold, AlphaTensor, AlphaGeometry and AlphaProof. We also discuss future research and the potential for such methods for ecological economic planning.

Despite all the industry hype and genuine advances, generative AI models are still prone to odd, inexplicable, and downright worrisome quirks. According to recent evidence, the industry's newer reasoning models may already possess the ability to manipulate and circumvent their human programmers' goals. Some AI will even attempt to cheat their way out of losing in games of chess. This poor sportsmanship is documented in a preprint study from Palisade Research, an organization focused on risk assessments of emerging AI systems. While supercomputers--most famously IBM's Deep Blue--have long surpassed the world's best human chess players, generative AI still lags behind due to their underlying programming parameters.

Researchers from the AI research organization Palisade Research instructed seven large language models to play hundreds of games of chess against Stockfish, a powerful open-source chess engine. The group included OpenAI's o1-preview and DeepSeek's R1 reasoning models, both of which are trained to solve complex problems by breaking them down into stages. The research suggests that the more sophisticated the AI model, the more likely it is to spontaneously try to "hack" the game in an attempt to beat its opponent. For example, it might run another copy of Stockfish to steal its moves, try to replace the chess engine with a much less proficient chess program, or overwrite the chess board to take control and delete its opponent's pieces. Older, less powerful models such as GPT-4o would do this kind of thing only after explicit nudging from the team.

Complex games like chess and Go have long been used to test AI models' capabilities. But while IBM's Deep Blue defeated reigning world chess champion Garry Kasparov in the 1990s by playing by the rules, today's advanced AI models like OpenAI's o1-preview are less scrupulous. When sensing defeat in a match against a skilled chess bot, they don't always concede, instead sometimes opting to cheat by hacking their opponent so that the bot automatically forfeits the game. That is the finding of a new study from Palisade Research, shared exclusively with TIME ahead of its publication on Feb. 19, which evaluated seven state-of-the-art AI models for their propensity to hack. While slightly older AI models like OpenAI's GPT-4o and Anthropic's Claude Sonnet 3.5 needed to be prompted by researchers to attempt such tricks, o1-preview and DeepSeek R1 pursued the exploit on their own, indicating that AI systems may develop deceptive or manipulative strategies without explicit instruction.

We demonstrate LLM agent specification gamnull ing by instructing models to win against a chess engine. We find reasoning models like o1null preview and DeepSeeknullR1 will often hack the benchmark by default, while language models like GPT null4o and Claude 3.5 Sonnet need to be told that normal play won't work to hack. We improve upon prior work like ( Hubinger et al., 2024; Meinke et al., 2024; Weij et al., 2024) by using realistic task prompts and avoiding excess nudging. Our results suggest reasoning models may resort to hacking to solve difficult problems, as observed in OpenAI (2024) 's o1 Docker escape during cyber capabilities testing.

Chess has long been a testbed for AI's quest to match human intelligence, and in recent years, chess AI systems have surpassed the strongest humans at the game. However, these systems are not human-aligned; they are unable to match the skill levels of all human partners or model human-like behaviors beyond piece movement. In this paper, we introduce Allie, a chess-playing AI designed to bridge the gap between artificial and human intelligence in this classic game. Allie is trained on log sequences of real chess games to model the behaviors of human chess players across the skill spectrum, including non-move behaviors such as pondering times and resignations In offline evaluations, we find that Allie exhibits humanlike behavior: it outperforms the existing state-of-the-art in human chess move prediction and "ponders" at critical positions. The model learns to reliably assign reward at each game state, which can be used at inference as a reward function in a novel time-adaptive Monte-Carlo tree search (MCTS) procedure, where the amount of search depends on how long humans would think in the same positions. Adaptive search enables remarkable skill calibration; in a large-scale online evaluation against players with ratings from 1000 to 2600 Elo, our adaptive search method leads to a skill gap of only 49 Elo on average, substantially outperforming search-free and standard MCTS baselines. Against grandmaster-level (2500 Elo) opponents, Allie with adaptive search exhibits the strength of a fellow grandmaster, all while learning exclusively from humans.