Self-play is an unsupervised training procedure which enables the reinforcement learning agents to explore the environment without requiring any external rewards. We augment the self-play setting by providing an external memory where the agent can store experience from the previous tasks. This enables the agent to come…
On-device research index
arXiv research
A locally-built, LLM-digested index of recent arXiv papers in quant finance, geometry/topology, and statistical ML — keyword search served straight from SQLite on this machine.
168,695 papers · 148 categories
Trend · papers per month
48 results for “Self-Play Learning”
Automatically finds effective security strategies through reinforcement learning and self-play.
problem Finding effective security strategies for intrusion prevention.
method Modeling interaction as a Markov game, evolving attack and defense strategies through reinforcement learning and self-play.
result Effective security strategies emerge from self-play, reflecting common-sense knowledge.
Truthful Self-Playstat.ML
Framework improves self-play for cooperative multi-agent learning.
problem Evolutionary learning converges to bad local optima in multi-agent RL.
method Add imaginary rewards using peer prediction method to elicit truthful signals.
result State-of-the-art performance in predator prey, traffic junction and StarCraft tasks.
This paper improves self-play learning in games by manipulating experience distributions.
problem Improving self-play learning in games through better experience sampling.
method Three approaches: weighted sampling, Prioritized Experience Replay, and diversifying trajectories.
result Major improvements in early training performance in some games, minor improvements overall.
Self-play, where the algorithm learns by playing against itself without requiring any direct supervision, has become the new weapon in modern Reinforcement Learning (RL) for achieving superhuman performance in practice. However, the majority of exisiting theory in reinforcement learning only applies to the setting wher…
New algorithm improves self-play reinforcement learning for competitive games.
problem Inefficient opponent selection in self-play reinforcement learning.
method Intelligently selects opponents based on adversarial rules derived from saddle point optimization.
result Algorithm converges to approximate equilibrium with high probability in convex-concave games.
The paper develops algorithms for competitive RL in partially observable MGs.
problem Challenges in reinforcement learning with function approximation and partial observability.
method Proposes posterior sampling methods for self-play and adversarial learning in zero-sum MGs.
result Developed algorithms achieve low regret bounds scaling sublinearly with GEC and episode number.
A promising approach for teaching artificial agents to use natural language involves using human-in-the-loop training. However, recent work suggests that current machine learning methods are too data inefficient to be trained in this way from scratch. In this paper, we investigate the relationship between two categorie…
Adversarial self-play in two-player games has delivered impressive results when used with reinforcement learning algorithms that combine deep neural networks and tree search. Algorithms like AlphaZero and Expert Iteration learn tabula-rasa, producing highly informative training data on the fly. However, the self-play t…
By introducing several improvements to the AlphaZero process and architecture, we greatly accelerate self-play learning in Go, achieving a 50x reduction in computation over comparable methods. Like AlphaZero and replications such as ELF OpenGo and Leela Zero, our bot KataGo only learns from neural-net-guided Monte Carl…
Paper optimizes reinforcement learning in self-play games with reduced steps.
problem Optimizing reinforcement learning algorithms for self-play in two-player zero-sum games.
method Proposes optimistic Nash Q-learning and Nash V-learning algorithms with improved sample complexity.
result Achieves sample complexity of for Nash Q-learning and for Nash V-learning, closing the gap with lower bounds.
SPIN converts weak LLMs to strong ones using self-play.
problem Growing strong LLMs without human-annotated data.
method Self-play fine-tuning (SPIN) starting from a supervised fine-tuned model.
result SPIN significantly improves LLM performance across benchmarks.
Goal-directed Reinforcement Learning (RL) traditionally considers an agent interacting with an environment, prescribing a real-valued reward to an agent proportional to the completion of some goal. Goal-directed RL has seen large gains in sample efficiency, due to the ease of reusing or generating new experience by pro…
Self-play fine-tuning improves diffusion models for text-to-image generation.
problem Plateauing performance of diffusion models after data saturation.
method Self-play fine-tuning (SPIN-Diffusion) using competition among model versions.
result Significantly improved model performance and human preference alignment.
In hierarchical reinforcement learning a major challenge is determining appropriate low-level policies. We propose an unsupervised learning scheme, based on asymmetric self-play from Sukhbaatar et al. (2018), that automatically learns a good representation of sub-goals in the environment and a low-level policy that can…
Despite the notable successes in video games such as Atari 2600, current AI is yet to defeat human champions in the domain of real-time strategy (RTS) games. One of the reasons is that an RTS game is a multi-agent game, in which single-agent reinforcement learning methods cannot simply be applied because the environmen…
The AlphaZero algorithm for the learning of strategy games via self-play, which has produced superhuman ability in the games of Go, chess, and shogi, uses a quantitative reward function for game outcomes, requiring the users of the algorithm to explicitly balance different components of the reward against each other, s…
Study learns optimal strategies in imperfect information games with self-play.
problem Learning optimal strategies in imperfect information games.
method Proposes Follow the Regularized Leader (FTRL) algorithms for imperfect information games.
result Proposes two FTRL algorithms: Balanced FTRL and Adaptive FTRL.
We introduce a new virtual environment for simulating a card game known as "Big 2". This is a four-player game of imperfect information with a relatively complicated action space (being allowed to play 1,2,3,4 or 5 card combinations from an initial starting hand of 13 cards). As such it poses a challenge for many curre…
SPPO optimizes language model alignment by treating preferences as a game and achieving state-of-the-art performance.
problem Capturing intransitivity and irrationality in human preferences for accurate language model alignment.
method Self-play-based approach to identify Nash equilibrium policy through iterative policy updates.
result SPPO achieves state-of-the-art win-rate of 28.53% on AlpacaEval 2.0 without external supervision.
Many poker systems, whether created with heuristics or machine learning, rely on the probability of winning as a key input. However calculating the precise probability using combinatorics is an intractable problem, so instead we approximate it. Monte Carlo simulation is an effective technique that can be used to approx…
eva method improves RL models by adaptively creating prompts, boosting performance.
problem Fixed prompt distribution limits scalability of RL models post-training.
method Evolving Alignment via Asymmetric Self-Play (eva) approach.
result Significant performance boost on benchmarks, e.g. 51.6% to 60.1% win-rate.
Algorithm converges to Nash equilibria in competitive games.
problem Finding Nash equilibria in decentralized, competitive Markov games.
method Decentralized Optimistic Gradient Descent/Ascent with a critic.
result Converges to the set of Nash equilibria under self-play.
A new algorithm reduces memory and computational needs for reinforcement learning.
problem Memory and computational inefficiency in model-free reinforcement learning.
method Memory-Efficient Nash Q-Learning (ME-Nash-QL) for two-player zero-sum games.
result Proves ME-Nash-QL reduces space and sample complexity for tabular and long-horizon cases.
The problem of retrosynthetic planning can be framed as one player game, in which the chemist (or a computer program) works backwards from a molecular target to simpler starting materials though a series of choices regarding which reactions to perform. This game is challenging as the combinatorial space of possible cho…
This paper proposes using a linear function approximator, rather than a deep neural network (DNN), to bias a Monte Carlo tree search (MCTS) player for general games. This is unlikely to match the potential raw playing strength of DNNs, but has advantages in terms of generality, interpretability and resources (time and …
In recent years, state-of-the-art game-playing agents often involve policies that are trained in self-playing processes where Monte Carlo tree search (MCTS) algorithms and trained policies iteratively improve each other. The strongest results have been obtained when policies are trained to mimic the search behaviour of…
A new method for RLHF using proximal point Nash learning.
problem Capturing real human preferences in RLHF.
method Proximal point Nash learning, embedding self-play updates into a proximal point framework.
result High-probability last-iterate convergence for the combined method.
High-throughput 3D control training system achieves 100,000 FPS.
problem Lack of efficient, single-machine reinforcement learning systems.
method Sample Factory combines asynchronous sampling and off-policy correction.
result Achieves 100,000 FPS on 3D control problems without sacrificing sample efficiency.
While we would like agents that can coordinate with humans, current algorithms such as self-play and population-based training create agents that can coordinate with themselves. Agents that assume their partner to be optimal or similar to them can converge to coordination protocols that fail to understand and be unders…
Faster WIND accelerates iterative BOND for LLM alignment.
problem Iterative BOND is inefficient in practice due to sample and computation inefficiency.
method Unified game-theoretic connection to self-play alignment, WIND framework with efficient algorithms.
result WIND variant achieves superior sample efficiency and faster computation.
Mobile network that millions of people use every day is one of the most complex systems in the world. Optimization of mobile network to meet exploding customer demand and reduce capital/operation expenditures poses great challenges. Despite recent progress, application of deep reinforcement learning (DRL) to complex re…
Adapts agent strategies on-the-fly for better cross-play in cooperative settings.
problem Cross-play issues between self-play agents and unseen partners.
method Adapts agent strategies using posterior belief updates via Gibbs sampling.
result Achieves strong cross-play in the Hanabi game without prior knowledge of partners' strategies.
This work improves autonomous racing by creating diverse opponents and adapting risk.
problem Balancing performance and safety in autonomous racing environments.
method Developed a self-play method using replica-exchange Markov chain Monte Carlo for diverse opponents and a distributionally robust bandit optimization for adaptive risk adjustment.
result Demonstrated real-time motion-planning methods achieving speeds comparable to Formula One racecars.
ISP improves DNN uncertainty for better subgroup accuracy.
problem Improving accuracy-group robustness in deep neural networks.
method Introspective Self-play (ISP) adds an introspection task to improve bias-awareness and uncertainty.
result ISP improves the accuracy-group robustness trade-off of AL methods.
Automatic machine learning is an important problem in the forefront of machine learning. The strongest AutoML systems are based on neural networks, evolutionary algorithms, and Bayesian optimization. Recently AlphaD3M reached state-of-the-art results with an order of magnitude speedup using reinforcement learning with …
We propose a method for tackling catastrophic forgetting in deep reinforcement learning that is \textit{agnostic} to the timescale of changes in the distribution of experiences, does not require knowledge of task boundaries, and can adapt in \textit{continuously} changing environments. In our \textit{policy consolidati…
On April 13th, 2019, OpenAI Five became the first AI system to defeat the world champions at an esports game. The game of Dota 2 presents novel challenges for AI systems such as long time horizons, imperfect information, and complex, continuous state-action spaces, all challenges which will become increasingly central …
To improve efficiency and reduce failures in autonomous vehicles, research has focused on developing robust and safe learning methods that take into account disturbances in the environment. Existing literature in robust reinforcement learning poses the learning problem as a two player game between the autonomous system…
We present a simple model of a non-equilibrium self-organizing market where asset prices are partially driven by investment decisions of a bounded-rational agent. The agent acts in a stochastic market environment driven by various exogenous "alpha" signals, agent's own actions (via market impact), and noise. Unlike tra…
Improved model-based reinforcement learning for multi-agent Markov games.
problem Suboptimal sample complexity for model-based algorithms in multi-agent reinforcement learning.
method Optimistic Nash Value Iteration (Nash-VI) for two-player zero-sum Markov games.
result First model-based algorithm matching information-theoretic lower bound with improved sample complexity.
Recent work in reinforcement learning demonstrated that learning solely through self-play is not only possible, but could also result in novel strategies that humans never would have thought of. However, optimization methods cast as a game between two players require careful tuning to prevent suboptimal results. Hence,…
Building an open-domain conversational agent is a challenging problem. Current evaluation methods, mostly post-hoc judgments of static conversation, do not capture conversation quality in a realistic interactive context. In this paper, we investigate interactive human evaluation and provide evidence for its necessity; …
JPS improves joint policies for multi-agent collaboration in imperfect information games.
problem Learning good joint policies for multi-agent collaboration with imperfect information.
method Decomposes global changes to localized policy changes, iteratively improving joint policies without re-evaluating the entire game.
result JPS improves solutions provided by unilateral approaches and outperforms algorithms designed for collaborative policy learning.
Quantum hardware and quantum-inspired algorithms are becoming increasingly popular for combinatorial optimization. However, these algorithms may require careful hyperparameter tuning for each problem instance. We use a reinforcement learning agent in conjunction with a quantum-inspired algorithm to solve the Ising ener…
Model-Free Learning for Two-Player Zero-Sum Partially Observable Markov Games with Perfect Recallstat.ML
Algorithm learns NE in imperfect information games with imperfect feedback.
problem Learning Nash equilibrium in imperfect information games with bandit feedback.
method IXOMD algorithm for model-free learning with convergence rate.
result IXOMD achieves convergence rate to NE.
Algorithm identifies correct hypothesis from alternatives in bandit problems.
problem Efficiently identifying the correct hypothesis from a finite set of alternatives in structured stochastic multi-armed bandits.
method Frank-Wolfe Self-Play (FWSP) reformulates the game as a saddle-point problem, using a differential-inclusion argument to prove convergence.
result Convergence of the game value for best-arm identification in linear bandits, with uniform global convergence to the optimal value.
Optimization of parameterized policies for reinforcement learning (RL) is an important and challenging problem in artificial intelligence. Among the most common approaches are algorithms based on gradient ascent of a score function representing discounted return. In this paper, we examine the role of these policy gradi…