Imperfect-Information Games

Overview

These games model settings such as poker, negotiation, and pursuit-evasion, where players must act using incomplete observations and beliefs about hidden cards, types, or intentions. Researchers develop equilibrium concepts and solvers such as counterfactual regret minimization, alongside Bayesian and Stackelberg methods for reasoning about beliefs, commitments, and incentives. Reinforcement learning, search, and opponent modeling produce agents for large games, while benchmarks and variance-reduced evaluation help compare their strength and robustness. A central challenge is scaling methods without losing strategic guarantees, especially when agents must adapt to opponents, exploit mistakes, or remain safe under uncertainty.

Latest papers

96 papers7 papers in the last four weeks, down 30% on the four weeks before. 0.1% of all new papers.

Nov 19, 2026

Safe Subgame Resolving in Multiplayer Games

Poker and other games with hidden information are too large to solve in one pass. Strong systems compute a complete blueprint strategy offline and refine it in real time inside the reached subgame. In 2p0s games, a safety theorem guarantees that subgame resolving never increases how much an adversary can gain by exploiting the resolved strategy relative to the blueprint strategy. With three or more players acting independently, no such guarantee exists yet, and heuristic resolving can increase the profile's best-response gap. We prove the first conditional safety theorem for uncorrelated multiplayer subgame resolving, a post-hoc certificate for any delivered update. The best-response gap of the resolved profile exceeds that of the blueprint by at most four terms. Two concern the resolver: the subgame value the blueprint promised without realizing (AiSA_i^{S}) and the resolver's subgame solve error (εi\varepsilon_i). The other two are the opponents' best-response gains inside the subgame over their promised values (∑qMq\sum_q M_{q}) and a structural tax (ΘΘ), which measures how much the players' total utility can change. The tax is zero in constant-sum games, including every rake-free poker variant. Constructions attain each term with equality. On three-player Leduc poker and on Kuhn poker with up to six players, resolving without a certificate raises the gap beyond a small tolerance in 12% to 100% of the updates. The certificate flags every such update and accepts 41% (Kuhn) and 86% (Leduc) of the standard resolving updates. A resolver that keeps the blueprint whenever the certificate exceeds the tolerance delivers no harmful update. The certificate needs only one best-response pass per opponent inside the subgame, so real-time multiplayer solvers that refine a blueprint can attach it as a safety check. Code is available at https://github.com/lbn187/MultiPlayerSolver
Nov 16, 2026

Instrumenting Failure: Scoring Language Model Agents at Every Decision Node in Imperfect-Information Games

Agent evaluations report task success and rarely say what an agent did at each decision. We score every decision node of three 12B to 14B instruct-tuned language models (LMs) playing five zero-sum imperfect-information games against four opponents on eight properties from decision and game theory, five of them against a CFR+ equilibrium reference. At each node, the LM also states a belief about the opponent's hidden state in a separate call. Without memory and at temperature 0, stated beliefs sit 0.250 to 0.288 in total variation distance from the equilibrium posterior, and where the action responds to the state its agreement with that belief exceeds chance by -0.04 to 0.19. The two calls frame the opponent differently, so the result is a behavioural inconsistency and makes no claim about internal beliefs. Under one prompt and greedy decoding, the LM's action never varies with the state in 10 of the 38 cells where the reference varies its own. Memory lowers the constant-action share in 9 of 13 testable ladders at temperature 0; at temperature 0.2 the cells at 1 fall from 16 to 5 of the 46 informative cells, while the belief results hold. A gate withholds, with a stated reason, any belief score the run cannot support, and seven of the eight non-zero defect rates are attributable to the instrument, a representation, or the LMs.
Nov 5, 2026

Learning and Steering in Incomplete-Information Games

In many modern multi-agent systems, a mediator seeks to influence strategic decision dynamics using only outcome-level feedback and without knowing agents' utilities. In this work, we study how a mediator can learn an unknown game and steer adaptive agents toward outcomes that maximize her utility using only trajectory feedback. The mediator knows the game form and her own objective, but not the agents' utilities. Moreover, she commits to nonnegative payments and provides recommendations and contexts, but receives only a single realized terminal outcome rather than full-information feedback. We develop halving-type algorithms that learn utility differences in normal-form games and conditional continuation-value differences in extensive-form games, achieving a sample complexity of O~(ε−2)\widetilde O(\varepsilon^{-2}) that matches worst-case lower bounds up to logarithmic factors. We then combine these learning procedures with a Lagrangian formulation in an online Learn-Compute-Steer algorithm that interleaves learning, equilibrium computation, and steering toward mediator-optimal extensive-form correlated equilibria (EFCE). With high probability, the Learn-Compute-Steer algorithm achieves a directness gap and an optimality gap of O~(T−1/4)\widetilde O(T^{-1/4}), while time-average payments are of O~(T−1/8)\widetilde O(T^{-1/8}).
Nov 4, 2026

Polynomial-Time Computation of Extensive-Form Proper Equilibria in Two-Player Zero-Sum Games

We give a deterministic polynomial-time algorithm for computing an exact extensive-form proper equilibrium in a finite two-player zero-sum extensive-form game with perfect recall and rational data. Properness is understood in the agent normal form: an action at an information set is penalized relative to a better action while all other local behavioral choices, including the player's own future actions, remain fixed. Our algorithm assigns each information set a geometric permutahedron of allowable action probabilities. A scaled doubly stochastic representation embeds these restrictions in a polynomial-size sequence-form polytope; a saddle point then satisfies the local ε\varepsilon-properness inequalities. To obtain an exact equilibrium rather than a solution at an arbitrarily small numerical perturbation, we establish sign stability of a parametric linear system, construct a persistent feasible basis, and extract the limits of behavioral ratios. The resulting probabilities are rational and have polynomial encoding length, including at information sets whose realization probabilities vanish.
Oct 29, 2026

A Robust Approach to Belief Mismatch in Non-cooperative Dynamic Games

Game-theoretic models are increasingly used to describe multi-agent interactions in control and robotics applications. However, theoretical and computational complexity makes these models intractable in many dynamic, incomplete information scenarios. Recent work makes strong assumptions to reduce computational complexity: agents share sensor measurements with one another and maintain a common estimate of state uncertainty. Unfortunately, these assumptions enforce a degree of cooperation in generically noncooperative interactions. This paper takes the perspective of an ego agent seeking both computational tractability and robustness to unanticipated behavior of other agents due to mismatched state estimates from private---rather than shared---sensor measurements. To that end, we formulate the game theoretic planning problem in an augmented belief space with perturbed Bayesian dynamics that allows the ego to predict all agents' potentially divergent state estimates. These perturbations are chosen by a hypothetical adversary, ``nature,'' to reflect how non-ego agents' state estimates---based on measurements the ego agent will not see at runtime---may deviate from the ego's own belief. Ultimately, our approach can be understood in the spirit of classical H-infinity robust control, extended to noncooperative, stochastic dynamic games. We present a solver with polynomial-time iterations and conduct two experiments which show how robust belief space reasoning leads ego agents to proactively limit their future vulnerability to model misspecification of other agents' beliefs and actions, reducing performance degradation by over 48%.
Oct 29, 2026

How a group that cannot communicate knows itself by its enemy

Teams that cannot communicate seem stuck with what each member sees. We show that their opponent can tell them the rest. Silent hunters, jammed robots and traders who never meet all face an adversary that reacts to the group as a whole; every member can read that reaction, a tell, and learn what no member knows alone. One tell can do the work of a message from every member. The adversary would rather hide, but hiding interferes with its own task, and a team that commits in advance to how it will respond decides how much hiding is worth. A committed team never loses when its opponent cares more about its own goals, and does as well as a fully informed team exactly when the opponent's stake outweighs any gain from lying. Below that point the team learns more by trusting less: in a solved Gaussian model it obtains the truth by promising not to use all of it and keeps the cost the opponent would have paid to conceal; an exact test decides when this holds in finite games. How a team spreads its trust matters too. A lone uninformed member with a full-support prior needs a private coin, but with additive squared-error losses two interchangeable members who respond differently replace it, and a small discount on trust insures against misjudging the enemy. Information, on this view, is not only gathered or aggregated; it is produced by how a team commits to respond.
Oct 28, 2026

Inner-Learning: From Hand-Coded Beliefs to Declarative Knowledge in Trick-Taking Games

We present a declarative approach to belief-state modeling in trick-taking card games. Starting from a state-of-the-art Skat player, we automatically extract its hand-coded inference rules into a domain-specific language built on the primitive ¬holds(P,C)\neg\mathsf{holds}(P,C), denoting that player PP cannot hold any card in set CC. Round-trip code generation verifies that the extracted rules reproduce the original engine's intended behavior, up to three audited bugs in the original that the declarative engine corrects. We then grow the rule base through symmetric derivation and annotation of an expert strategy document---unifying code, symmetry, and expertise in one representation. We extend the knowledge representation with cardinality constraints: at-most-kk and at-least-kk bounds on the cards a player holds from a named group, expressed in the same declarative language and enforced at world-generation time. Confidence parameters connect symbolic rules to probabilistic belief updates. Self-play evaluation shows the declarative engine measurably improving winning percentage and score over the original hand-coded engine, with the extracted rules, the human-annotated rules, and the cardinality constraints each contributing to the effect.
Oct 21, 2026

When Are We Near Nash? Certifying Solutions in General-Sum Games

Heuristic solvers for general-sum extensive-form games produce a stream of iterates, and practice reports the NashGap of the returned profile. We ask what a solver can know about its own iterates from the regret accumulators and the terminal-distribution accumulators its updates already maintain. In two-player games the NashGap of a weighted average strategy equals, at the maximizing player, the weighted external regret plus a signed value correction. Bounding the regret by the running counterfactual-regret accumulator gives an online certificate whose cost per checkpoint is linear in the number of terminal histories. The certificate stays sound under alternating updates and in games with three or more players. The mismatch between what alternation evaluates and what the averages record enters as one signed comparator term, maximized in a single backward pass per player whose comparator weight is nonzero. The bound is read off the regret state, with no best response against the average. Run a portfolio of training-free regret-matching solvers, output the iterate with the smallest certificate, and verify that iterate once. On the benchmark's biased Leduc family a pool of sixteen training-free solvers, seven original and nine recent variants, solves 89%89\% of the 6464 instances at threshold 10−510^{-5} this way, matching the trajectory-minimum union of the solvers. The same protocol solves 92%92\% of the 6464 pre-registered held-out instances and 89%89\% again on a third set of 6464 instances that no earlier stage used. The zero-best-response certificate certifies all 5757 in-sample solves and all 5757 fresh ones, and 5858 of the 5959 held-out ones. Its exact-comparator refinement certifies every solve in the three sets. The certificate states both and attributes its bound. Code is available at https://github.com/lbn187/CertNE.
Oct 19, 2026

Teaching AI Agent to Play Poker Strategically Through Preference Optimization

A poker agent can make a legal move and still make a strategically poor one. We ask whether compact AI agents can exhibit strategic behavior under hidden information and whether preference optimization can improve beyond static and zero-shot policies. We first evaluate a static baseline and three frozen language models: SmallThinker, Qwen3, and Gemma-3 whose performance varies substantially between exact-game and six-player evaluation. We therefore propose Strategic Preference Optimization with KL-anchored Equilibrium (SPOKE), combining teacher-relative value penalties, policy anchoring, risk constraints, and structured rationale supervision, and compare it with SFT, DPO, GRPO, and GSPO. Across three runs of a SmallThinker-4B-A0.6B policy, SPOKE achieves the lowest mean exploitability in Kuhn (0.213) and Leduc (1.081) and the highest six-max top-1 agreement (0.775), while maintaining teacher KL comparable to SFT (0.522 vs 0.523). These results provide preliminary evidence that strategic preference optimization can improve strategic quality while preserving teacher-consistent behavior.
Oct 14, 2026

SafeAdapt-Poker: Counterfactual Risk Calibration for Budgeted Adaptation under Opponent Shift

Opponent histories can mislead after behavioral shifts. SafeAdapt-Poker formulates budgeted adaptation under opponent shift as a dual-counterfactual sequential-control problem using only information available before each decision. At each decision the controller asks two distinct questions: what local downside a deviation from the frozen anchor would induce under matched legal continuations, and what future adaptation value would be lost by consuming part of the remaining deviation budget now. The first question is answered by a monotone strategic-risk model over deviation intensity; the second by a paired budget-state continuation model that prices current budget consumption. The controller selects among support-preserving candidate deviations by trading immediate adaptation value against these two quantities, a pathwise cumulative-KL ledger bounds total anchor-relative deviation, and a finite family of frozen controllers is screened on complete paired calibration sessions before held-out interaction. A finite-candidate perturbation bound shows that the allocation objective degrades continuously with continuation-value estimation error. Evaluation covers exact-game exploitation and safety, four opponent shifts, local-risk calibration, delayed-opportunity allocation, direct opportunity-value estimation, value-error robustness, factorial mechanism attribution, budget sensitivity, abstract-HUNL cross-play, and player/time-held-out public-history behavior. On exact Leduc the full controller attains a matched weak-pool gain of +0.0718 with a worst-opponent delta of -0.0126, raises delayed-opportunity capture from 0.6318 under myopic allocation to 0.7841 under the same session budget, and reaches +42.7 mbb/hand in abstract-HUNL cross-play; the retained public-history screen reports 0.5820 action accuracy and 0.9277 log loss versus 0.5039 and 0.9778 for a public prior.
Oct 13, 2026

IR-PRM+ with Previous-Utility Prediction: Constant Regret and Best-Iterate Convergence

Regret matching (and its variants) and counterfactual regret minimization are central algorithms in practice for normal-form and extensive-form game solving. Yet these have historically lacked the stronger theoretical convergence guarantees available to optimistic and extra-gradient methods. Recent work introduced IREG-PRM+, a regret matching variant which does have strong theoretical convergence guarantees. However, it relies on a simultaneous, extra-gradient update. IREG-PRM+ is an instance of the more general IR-PRM framework, where empirically optimistic and alternating updates often have better performance. However, the prior extra-gradient analysis does not extend to these cases. We introduce a new approach that covers both settings. For simultaneous and alternating IR-PRM+ in finite two-player zero-sum matrix games, previous-utility prediction recovers the constant-regret and convergence-rate guarantees previously obtained with extra-gradient prediction: external regret bounded independently of T, an O(T^{-1}) average duality gap, and an O(T^{-1/2}) best-current-iterate gap. Moreover, the current duality gap vanishes and the current profile converges to the Nash equilibrium set. The key step bounds the component of consecutive strategy movement not controlled by the existing IR-PRM+ regret inequality using the relative growth of the nondecreasing regret-vector norm rho_t. Coupling this bound across the two players yields recurrences whose effective coefficient scales as 1/(rho_t^X rho_t^Y), preventing unbounded norm growth. We extend the analysis to finitely many simplex learners under a weighted Minty condition and, through the corresponding CFR variant IR-PCFR+, recover analogous guarantees for finite two-player zero-sum perfect-recall extensive-form games under the same full-reach condition as prior work.
Oct 13, 2026

On Symmetry for Characterizing Agent Heterogeneity in Multi-Agent Systems

Understanding agent heterogeneity can inform the design of shared and agent-specific policy representations, however existing criteria provide no unified structural test. We study finite-horizon POSGs and define a type-compatible transformation as a policy-evaluation symmetry when complete policy evaluation commutes with it for every decentralized behavioral policy. Policy-evaluation symmetries form a group for which agent orbits define structural agent types. This classification does not depend on the particular policy chosen or the distribution of initial states. We prove that every componentwise environment automorphism is a policy-evaluation symmetry, but that the converse does not hold. Applying the framework to cooperative Dec-POMDPs shows that agents can belong to different structural types even when they share a common reward. For prescribed weighted reward aggregation, we derive an exact value identity and sufficient conditions for the resulting Dec-POMDP to inherit the policy-evaluation symmetries of the source POSG. Finally, we give sufficient conditions under which repeated model factorizations induce policy-evaluation symmetry, and show that preserving an optimal Boltzmann policy set neither implies nor follows from this symmetry in general.
Oct 7, 2026

The Confidence Game: Strategic Miscalibration in Human-AI Delegation

Calibrated uncertainty quantification is essential to ensuring AI agents are trustworthy and reliable. However, when agents seek to maximize user engagement or revenue, confidence reports may be strategically distorted, detracting from their informativeness. We formalize this problem in the Confidence Game: a repeated signaling game with imperfect monitoring in which an agent of unknown honesty and ability reports its confidence, and a user decides whether to delegate the task or complete it herself. The agent manages the tradeoff between manipulating signals and maintaining its reputation. We characterize the Markov Perfect Bayesian Equilibria of the two-period game and show that honest reporting is not an equilibrium, inflation is the unique best response once the agent is sufficiently myopic, and under-reporting requires that the user believe honesty to be a minority. We then place an LLM in the agent role, supplying it with its true probability of success so that any gap between what it knows and what it reports is attributable to incentives rather than to miscalibration. The model claims high confidence on 56% of tasks it has been told it will probably fail. This persists on real tasks, where it must estimate its own accuracy and causes miscalibration to increase while the agent's signal becomes less informative. Furthermore, we find that the LLM agent's decisions are coherent, but it systematically underestimates both how likely the user is to delegate and how secure its reputation is, resulting in less extreme behavior. Pricing the agent's reporting rule, we find that it destroys 68% of the gains from delegation, of which 71% is information the report no longer carries and no amount of user sophistication recovers. Overall, we establish confidence reporting under delegation as a strategic problem and provide a tractable basis for modeling, analyzing, and testing agent behavior.
Oct 6, 2026

BluffJAX: Adversarial Imperfect Information Games in JAX

We introduce BluffJAX: an open-source suite of adversarial imperfect information games in JAX. We provide canonical implementations of games designed for high simulation throughputs and parallelization on GPU accelerators. Our suite consists of well-studied benchmarks such as Texas Hold'Em Poker and Kuhn Poker, as well as games that have not been previously studied in reinforcement learning research, such as Bluff, Stud Poker, and Kemps. We hope that implementing a variety of game mechanics and difficulties will introduce new challenges and foster novel research directions in game-theoretic methods for RL. We benchmark the throughput performance and memory usage of our environments in single and multi-GPU settings, demonstrating scaling of up to hundreds of millions of samples per second, and motivating the usage of BluffJAX over related GPU and CPU-based libraries. We benchmark reinforcement learning, tree search, and game-solving algorithms in JAX in order to provide users with baseline results and facilitate future comparisons.
Oct 4, 2026

Communication Shapes Collective Inference in Self-Adapting LLM Societies: Evidence from Mafia

When does communication help a group identify hidden adversaries, and how does its value change as the group adapts? In Mafia, an informed minority hides inside an uninformed majority whose only evidence is open play. The zero-information game, where each day's vote eliminates a random player, is exactly solved and scores every society; matched-casting comparisons between protocols identify the effect of communication. Societies of 8-100 claude-haiku-4-5 agents (7,416 analyzed games, 1.9M model calls) adapt by rewriting and inheriting private strategy notes. Simultaneous broadcast improves adversary identification over silence in all nine compositions tested (8-46 players). Turn-taking removes most of this advantage; its voting landslides are as frequent as broadcast's but land on mafia near chance (1.08x versus 2.53x). At 70 players, agents reading eight statements per day identify adversaries worse than silent ones, and limited talk is worth less than at 46 players. Adaptation is fast but need not help. In their first broadcast games, citizens announce their role far more often than mafia (91% vs. 30%) and first-day votes find mafia at three times chance; within two generations citizens stop announcing and the cue fades, a change the inherited notes carry. In controlled redeployments at 16 players, societies carrying sixty generations of their own notes score below societies with none. Communication shapes both collective inference and the signals it depends on, so a protocol's value must be measured together with the adaptation that changes those signals.
Oct 4, 2026

Pessimistic Minimax Learning for Public-Private Information Games under Unilateral Coverage

We study offline learning in two-player zero-sum contextual games with public and private information, motivated by strategic settings such as auctions and negotiations with private valuations. We introduce unilateral prescriptive concentrability and show that asymmetric information can change offline coverage through its effect on equilibrium behavior. For finite state-action spaces, we develop a pessimistic algorithm with an O~(1/n)\tilde{O}(1/\sqrt{n}) exploitability rate, matching the standard sample-size dependence for fully observed minimax games. We further develop a pessimistic policy mirror descent framework, PPA-PMD, for general function approximation and obtain a unified O~(1/n+1/T)\tilde{O}(1/\sqrt{n} + 1/\sqrt{T}) exploitability rate with no-regret actor updates. Together, these results provide the first theoretical framework for offline equilibrium learning under public-private information constraints.
Oct 4, 2026

Orchestrating Level-KK Policies Against Unknown Opponents in Partially-Observable Dynamic Games

Level-KK reasoning generates a hierarchy of policies specialized to opponents with different reasoning levels. When an opponent's level is unknown, a common deployment rule estimates that level and selects the corresponding response. In a dynamic, partially-observed game, this selection is repeated, with each choice shaping subsequent states and observations. The response associated with the most likely opponent level need not maximize expected return from the current history. We formulate this deployment problem as dynamic orchestration of a fixed, pretrained policy library in a partially observable Markov game. We compare classification-based orchestrators (CBOs) trained using offline data or on-policy data aggregation with a reinforcement-learning-based orchestrator (RLBO) trained to maximize expected discounted return. In pursuit-evasion experiments, on-policy training improves classification and return, yet RLBO achieves higher return than the on-policy and offline CBOs. Given a pretrained library, RLBO also reaches performance comparable to a policy trained directly over the pursuers' action space with fewer training timesteps. These findings support treating a hierarchy of level-KK policies as a resource for orchestration, not a prescription for deployment.
Sep 28, 2026

SAGE: Structured Strategic Reasoning for Efficient LLM Game Playing

A strong LLM strategic agent should reason prospectively over uncertain futures, adapt its strategy to opponents' behavioral tendencies, and continuously recalibrate its decision process from interaction experience. However, incorporating these sources in free-form reasoning could lead to unsupported strategic assumptions, inconsistent opponent estimates, and harmful interference from irrelevant historical interactions. To address these issues, we propose SAGE, a training-free inference-time framework that structures LLM strategic reasoning around three coordinated operations: anchor, adapt, and recalibrate. SAGE first anchors reasoning to an equilibrium policy that provides a strategically valid prior. It then conditions deviations from this anchor on a soft belief over opponent behavioral tendencies, enabling opponent-specific exploitation. Finally, SAGE distills strategically related interactions into counterfactual hypotheses about previously missing considerations, allowing past experience to recalibrate the model's reasoning. We evaluate SAGE on three repeated imperfect-information games: Leduc Hold'em, Liar's Dice, and Goofspiel, against various opponent types in each game. Compared with reasoning-intensive LLM agents, including Suspicion-Agent, ReTA, Agent-Pro, EMO, and Hypothetical Minds, SAGE achieves up to a 127.6% payoff improvement in Liar's Dice while reducing input and output token usage by up to 80% and 90%, respectively. In direct match-up play, it attains non-negative mean payoff against 5/10, 8/10, and 8/10 evaluated opponents in Leduc Hold'em, Liar's Dice, and Goofspiel, respectively, while using relatively fewer tokens. Code is available at https://github.com/chenzhwsysu57/SAGE.
Sep 27, 2026

RSD-Poker: Structure-Adaptive and Shift-Robust Risk-Utility Certification for Residual Policies in Imperfect-Information Games

Residual policy adaptation provides a lightweight way to modify a strong reference policy, but a shared scale and a fixed subgroup partition can hide heterogeneous degradation and become fragile when the deployment mixture of information states changes. We introduce RSD-Poker, a structure-adaptive and shift-robust certification framework that freezes a bank of residual families and scales, learns a policy-visible partition on an independent structure split, and freezes that partition before calibration labels are joined. Each candidate-group pair receives a weighted simultaneous upper certificate for anchor-relative risk and a lower certificate for weak-response utility. A robust group-to-candidate map is then selected over a predeclared uncertainty set of deployment group proportions. Under independent calibration units drawn from each frozen group's law, a candidate bank and partition fixed before calibration, and invariant within-group conditionals, the selected map satisfies its declared mixture-robust risk budget and utility certificate with probability at least 1−ζrisk−ζutil1-ζ_{risk}-ζ_{util}. The information contract supports both a teacher-backed transform and a teacher-free observation-only student. The retained deterministic 24-state audit remains an exact replay diagnostic: empirical-zero selects α=0.08α=0.08, raising the weak-response proxy from 4.2082 to 4.2889 with 0/120/12 held-out threshold crossings. On stratified held-out states, the learned-partition dual selector raises weak utility from 4.4074 under global dual certification to 4.4936 and lowers held-out violation from 0.0215 to 0.0078; its mixture-robust variant reaches violation 0.0059. Across five observation-only checkpoints, risk-calibrated residuals attain weak utility 4.3659±0.01774.3659\pm0.0177 and violation rate 0.0178±0.00570.0178\pm0.0057.
Sep 10, 2026

GPU-CFR: 80x Faster Counterfactual Regret Minimization by Compiling the Game to Static Dataflow and CUDA Graph Replay

Counterfactual regret minimization (CFR) is one of the few large numerical workloads that still runs faster on CPUs than on GPUs. Each iteration sweeps a game tree with up to billions of states in millions of small, interdependent gather and scatter steps issued through a generic tree interface. On a GPU every kernel finishes in microseconds, so kernel launches and framework dispatch dominate the run time, and prior GPU implementations have lost to optimized CPU code. We observe that for a fixed game, everything about a CFR iteration except the numerical values is known before the first iteration runs. We propose GPU-CFR, a compiler and runtime built on this observation. It compiles any game once into static dataflow: flat edge and information-set arrays, precomputed indices, and depth-level batched passes fix the entire operation sequence, and only solver state changes between iterations. Static chance folding, depth-level execution blocks, and a dual-lane reach buffer cut the number of framework operations by up to 18.1x. Because shapes, indices, and buffer addresses never change, CUDA Graph Replay records the iteration once and replays it with a single graph launch. On one A100, across an eight-game suite that spans card games, dice games, and board games, GPU-CFR runs 29.8--80.4x faster than the fastest prior GPU CFR on the same accelerator, and 14--258x faster than LiteEFG, one of the fastest open-source CPU implementations, on the four largest games. The compiled representation carries most of that margin: on eight CPU threads with no accelerator it is already 2.2--51.1x faster than the GPU baseline. On the CPU the optimized path reproduces the reference iterates bitwise, and tree construction and graph capture pay for themselves within the first solve. GPU-CFR beats every CPU and GPU baseline on the mid-to-large games of the suite without changing the update rule.
Sep 7, 2026

Deception in Reach-Avoid Game with Unknown Heterogeneous Attackers Speed Information

This letter investigates a reach-avoid game involving two Attackers and one Defender, where the Attackers aim to maximize the number reaching the target region while the Defender seeks to minimize it. In contrast to conventional complete information formulations, we consider an information asymmetry scenario where the Attackers' heterogeneous maximum speeds are privately known but publicly disclosed to lie within continuous ranges. Existing studies on uncertain speeds, however, have primarily focused on homogeneous settings, whereas heterogeneity extends the uncertainty from a common capability level to the relative capability configuration of the Attackers. To address the resulting capture-order ambiguity over infinitely many possible speed combinations, we establish a critical speed pair framework that characterizes when different capability configurations induce different optimal capture orders, and enables the analysis of the Defender's guessing behavior and the design of information-limiting strategies for the Attackers. We demonstrate that under certain initial conditions, the Attackers can mislead the Defender into making suboptimal decisions through a slow-speed deception strategy, achieving superior payoffs compared to the complete information game. Numerical visualizations reveal the widespread occurrence of such dilemma conditions.
Sep 6, 2026

Unsound Search with Policy and Value Networks in Legends of Code and Magic

Decision-time search in perfect and imperfect information games with enumerable belief states are effective methods for game AI. Collectible card games are imperfect information games with large belief states. Legends of Code and Magic is a collectible card game competition where the belief states are 21012^{101}. The Legends of Code and Magic (LoCM) champion, ByteRL, plays with no search. Other works claim sound enumeration-based search is unusable in the genre due to the number of belief states. We measured three previously defined properties that predict where theoretically unsound perfect information Monte Carlo's defects are cheap and found LoCM sits in the favorable region. Starting with imitation learning of the runner-up policy, NeteaseOPD, we created a policy and value feed-forward network. Our agent searches over worlds sampled from a prior over the opponent's deck built from the runner-up's drafts. Using our strictest configuration in the battle phase we beat ByteRL with a win percentage of 51.35% 95% CI [50.37, 52.33], over 10,000 pre-registered games using the LoCM official referee and time limit. Search is not a minor factor on the matchup between our agent and ByteRL. Without search this agent scores 26.8% and adding search adds +24.6 points. Unsound search in imperfect information games could be exploitable. We replicate a published best-response attack against ByteRL. We then apply the same attack protocol to two search configurations of our agent, and each one resists it better than ByteRL at every iteration. In LoCM unsound search gives us a stronger and more resilient agent.
Sep 3, 2026

Abstraction Agent

Information abstraction, which groups strategically similar private states into a tractable number of buckets, is essential for scaling game-solving algorithms to large imperfect-information games. Constructing effective abstractions, however, has traditionally required domain-specific evaluators such as hand-strength calculators or equity estimators, which demand expert knowledge and engineering effort and are unavailable for most less-studied games. We propose the Abstraction Agent, a zero-shot pipeline that uses a large language model (LLM) to discover continuous strategic features from a natural-language game description, score private states on these features, and cluster them into abstraction buckets, without any game-specific evaluator, training data, or game-tree traversal during abstraction construction. The pipeline runs in four phases: feature discovery with calibration anchors, batched private-state scoring, correlation-based feature selection, and kk-means clustering. The resulting abstractions reduce lifted-strategy exploitability by up to 62% relative to an expected-hand-strength baseline on heads-up no-limit Texas hold'em (HUNL) turn endgames, and beat a scalar rank baseline at every granularity on ROVER Trials, an original game absent from any pretraining corpus. Beyond these quantitative benchmarks, the pipeline transfers with unchanged prompts to four-card Pot-Limit Omaha, HUNL preflop and flop, and Riichi Mahjong, where the discovered features track each game's recognized strategic concepts. This is structured knowledge elicitation: converting implicit strategic knowledge in LLM parameters into explicit numerical features for downstream algorithmic computation. The code is available at https://github.com/lbn187/AbstractionAgent.
Sep 1, 2026

NashDreamer: Model-Based Reinforcement Learning for Zero-Sum Imperfect-Information Games

Model-based reinforcement learning (MBRL) has achieved remarkable results in single-agent domains, yet its extension to competitive imperfect information games (IIGs) remains underexplored. In multi-agent settings, opponent-induced non-stationarity complicates the learning process, and decentralized model learning faces severe identifiability barriers, which we argue make centralized model learning a mathematical necessity. Building on this analysis, we propose NashDreamer, a principled MBRL framework for two-player zero-sum IIGs. It introduces a centralized Multi-Agent Recurrent State-Space Model (MARSSM) that decouples environment dynamics from the effect of players' strategies on their individual observations. NashDreamer is designed to use arbitrary policy gradient algorithms and inherits their convergence guarantees towards Nash equilibria under an idealized model. Empirical evaluations across five benchmark games demonstrate that NashDreamer substantially improves sample efficiency over model-free baselines early in the training. Finally, we theoretically analyze the architecture's optimization landscape, identifying the vulnerability of the Dreamer family of algorithms to posterior collapse in stochastic environments. We leave it as an open challenge.
Aug 16, 2026

CoupVisor: Strategy Optimization by Round and Challenge Decision Support

This paper presents CoupVisor, a decision-support system for the hidden-information card game Coup. It addresses two questions: what a player should do on each turn, and when a player should challenge an opponent's claim. The system is built around a single description of game events, which is shared across manual play, replay of recorded games, simulation, belief tracking, advisor recommendations, and learning-based policies. CoupVisor estimates the chance that a claim is truthful by combining how likely each role is with how many cards the claimant still holds, which corrects a case where the very first claim of a game was flagged as suspicious despite no evidence. We compare a rule-following advisor and several learned and heuristic players across many simulated games and different opponent styles. Our main finding is that the choice of reward, whether it rewards short-term gains or ultimately winning the game, decides which learning approach performs best, and that a win-oriented reward produces a policy that outperforms all baselines.
Aug 15, 2026

Do Geometry-Aware Positional Encodings Help Transformers in Spatial Imperfect-Information Games?

Transformers applied to spatial imperfect-information games must represent map geometry while tracking hidden entities through time. We ask whether geometry-aware positional encodings improve these capabilities, without claiming a new positional encoding. We construct a four-level benchmark on a hexagonal naval pursuit game: controlled geometry and topology probes, an exact-Bayes hidden-target tracking task, offline policy imitation at 1k and 10k games, and 7,200 fixed-seed games against three legacy opponents. Across matched Transformer backbones, HexRoPE reduces exact-belief posterior cross-entropy relative to no positional encoding by 0.278 on D6-transformed test orbits and 0.329 on a larger map; both hierarchical-bootstrap confidence intervals exclude zero, and both Holm-adjusted p-values are below 0.001. At 1k games, HexRoPE improves policy action accuracy by 4.63 percentage points over no encoding and 2.05 points over rectangular relative bias; the gains shrink to 1.55 and 0.41 points at 10k games. However, HexRoPE does not improve aggregate gameplay win rate: its paired effect over no encoding is -1.56 percentage points (95% CI [-4.50, 1.17]). Rectangular relative bias is strongest on D6 belief consistency but fails sharply when extrapolating from radius 3 to radius 4, while graph bias provides only a small blocked-edge gain. The results show that geometric inductive bias improves belief estimation and data-efficient imitation, but those representation gains do not automatically produce stronger closed-loop play.
Aug 14, 2026

CFR without Unbiasedness: Debit-Budgeted Public Feedback

At a finite public-chance cut, partial counterfactual regret minimization (CFR) can evaluate successive outcome batches at learner states changed by earlier batches. Balanced coverage then lacks the conditional unbiasedness used by Monte Carlo CFR analyses. We introduce target debit, the one-sided cumulative underestimation needed to transfer regret on delivered feedback to the dense target game. Our central theorem bounds weighted exploitability by delivered-row regret, target debit, and a native alternating-update correction. This yields Debit-Budgeted CFR (DB-CFR), which selects a public plan from completed history, executes it once, and settles its debit afterward. Complete coverage has zero debit and supplies exact recovery. With at most ℓ\ell unsettled blocks, hidden debit is bounded by the sum of the ℓ\ell largest predictable block envelopes; for fixed ℓ\ell and polynomial output weights, the resulting O(T−1)O(T^{-1}) term preserves the additive-RM/RM+ rate. A finite converse gives equal-metadata plans with charges 00 and 11/811/8, proving that a nontrivial strict-budget decision requires semantic information or bounded post-execution settlement. In two released heads-up no-limit hold'em turn endgames, all 42 registered shallow comparisons favor partial coverage before complete coverage overtakes it between 32 and 64 rounds. In public Leduc, a lagged rule fixed on four development orders beats the best matched-recurrence-work fixed width on all ten unseen orders (median NashConv gain 0.04560.0456) without future candidate replay.
Aug 10, 2026

ICM Out! Better Tournament Strategy from Computed Continuations, vs. Solvers and LLMs

The Independent Chip Model (ICM) converts tournament chips into reference prize equity, and policies are routinely constructed against those values. Because ICM reads only stack sizes, it omits action order, blind obligations, and seat rotation, and it does not price the elimination pressure a big stack puts on the short stacks it can bust. Those omissions can alter the successor-state contrasts that determine a move. We introduce Strategic-Continuation Optimization (SCO), a policy-construction method that enumerates current-hand outcomes, maps them to successor states, prices those states with continuation values computed from the finite tournament model, and optimizes and freezes the resulting current-hand policy. The fixed-ICM comparison policy changes one thing only: the same optimizer solves the same game with successor states priced by analytic ICM, so the two policies differ only through that pricing. We evaluate the resulting policies in a three-player jam/fold tournament with a $1M prize pool. Relative to the frozen strategic-continuation benchmark, analytic ICM has $9{,}036 mean absolute value error across all 2,838 state--seat entries. That value error rewrites the ranges it prices: measured against each decision point's own fixed-ICM jam range, SCO moves the jam frequency by an average of 14.08%. To price those different moves, we compare all 946 states and three policy owners while changing only the focal policy and holding both opponents and the continuation evaluator fixed. The policy produced by SCO earns $214.33 more prize equity per hand on average and is favored in 2,433 of 2,838 matched units. The ordering survives replacing the solver-built opponent with two LLMs and with a family of non-modeling threshold players. This value-to-policy-to-cost chain shows directly when ICM becomes an inadequate objective for tournament strategy construction.
Aug 10, 2026

Tracking the Best Strategy in an Extensive-Form Game

We consider the extensive-form bandit problem where on each trial the learner plays an extensive-form game against an oblivious adversary. We focus on the notion of switching regret, which measures the expected performance of the learner against that of any switching sequence of mixed strategies in retrospect. Our algorithm takes a parameter ρ>0ρ>0 and achieves a switching regret of O~((1/ρ+ρK)HAT)\tilde{\mathcal{O}}((1/ρ+ρK)\sqrt{H A T}) where KK is the number of switches in the comparator sequence, HH is the maximum number of the learner's information sets that can be traversed during a play of the game and AA is the number of actions that the learner can possibly take. Our algorithm is extremely efficient, taking a per trial time of only O(HB)\mathcal{O}(H B) where BB is the maximum number of actions available to the learner at any of its information sets.
Aug 7, 2026

Solver-Guided Reasoning for Mixed-Equilibrium Strategies

Reasoning in large language models (LLMs) is often grounded in human text, human demonstrations, and human-generated rationales. For equilibrium reasoning in complex games, however, relying on human data can be suboptimal. In fact, human play is often guided by intuition and heuristics and can deviate substantially from game equilibrium. This discrepancy is amplified in games with mixed-strategy equilibria, where human data is heavily biased toward pure strategies. Consequently, conditioning LLMs on this data yields weak game strategies. To grant LLMs the reasoning capacity in games, in this work, we study how to elicit equilibrium play using solver output. We propose Mixed-Strategy Decision Tree (MDT), which articulates the silent optimality of the equilibrium into sparse strategic rules that both humans and LLMs could understand. Using solver output rather than human annotation allows us to extend the input to arbitrarily new states and continuations. We instantiate this study on No-Limit Texas Hold'em by querying a solver oracle for over 250 million mixed-strategy decisions; MDT together with other techniques \textbf{reduces the ℓ1\ell_1 distance to the equilibrium by 52.6%52.6\%} across 88 different LLM configurations. A Route-only ablation tests the incremental contribution of the shadow-based contrast, while complete River-endgame and Liar's Dice experiments evaluate strategic fidelity and portability beyond the original NLH communication setting.