self-play

R^2Orch introduces a recursive self-play paradigm for training agent orchestrators with learnable task selection and a Skill Withdrawal outer loop, eliminating hand-designed curricula and outperforming baselines across twelve reasoning and interactive datasets.

R²Orch 提出递归自博弈编排范式，通过内层自博弈与外层技能撤除循环，实现智能体编排的递归自我改进，在12个数据集上超越基线。

An agent trained via self-play reinforcement learning on a single desktop (RTX 3090) achieves an 89% win rate against Warcraft III's hardest built-in AI, running the real unmodified game under Wine with a custom shim and harness, in 840,000 games over three weeks.

Researchers from MIT CSAIL stranded an AI agent named Eko, powered by Claude Code, on a virtual island to observe whether it could learn through free play without explicit instructions. Over 30 hours, the agents exhibited surprising emergent behaviors, exploring whether autonomous, self-directed learning is possible for AI systems.

微软研究院等机构提出RRI（Recursive Reward Improvement），通过引入对抗性奖励黑客代理来改进模型生成评分标准（rubric）的质量，在HealthBench Professional上显著优于仅生成任务的自改进基线，推动了开放-ended专家任务的自博弈训练方法。

SHRED 是一个通过大规模异构自博弈（车辆、行人、骑行者）从随机初始化联合训练的框架，无需人类驾驶演示或日志轨迹回放，即可学习真实且交互性强的交通行为。在 Waymo Open Sim Agents Challenge 和 InterPlan 基准上，SHRED 优于依赖演示的自博弈方法和基线规划器。

The paper introduces Modality Self-Play (MSP), a training framework that enables multimodal foundation models to compose across modalities without real paired cross-modal data, using proxy textual descriptions as semantic bridges and achieving state-of-the-art or competitive results on multiple multimodal benchmarks.

这篇名为CoSIR的预印本提出通过可验证信息奖励（基于贝叶斯惊异度降低）对LLM进行自博弈强化学习后训练，使模型显式学习多智能体协作能力，并在7个未见过的协作基准上取得最优表现，同时基本保持原有单任务性能。

Tengyu Ma highlights AC2 (Actor-Critic with Action Chunking), a reinforcement learning method that uses a learned critic to score token chunks, enabling partial rollouts and faster training than GRPO for LLMs.

提出GGSD框架，通过自博弈游戏引导技能发现，让人类可用少量离散技能直接控制具身智能体（Ant、Franka臂、Unitree G1），并在无需额外训练的情况下组合技能解决未见过的任务。

The paper introduces Self-Play Search Distillation (SPSD), a framework that uses self-play of MuZero-like networks to generate superhuman synthetic data for training LLMs, significantly improving reasoning performance on mathematics benchmarks.

The tweet introduces a research concept called Self-Play Pretraining with Zero Data, where two models start from random initialization and generate pretraining data through self-play on a universal Turing machine.

The article introduces Self-Play Pretraining with Zero Data, a research method where two LLMs start from random initialization and learn through a generator proposing programs for a universal Turing machine, demonstrating how inductive biases can enable general learning.

This paper introduces Self-Play Search Distillation (SPSD), a framework that uses self-play in board games to generate synthetic data for improving large language model reasoning, with demonstrated improvements on mathematical benchmarks.

Skild AI demonstrated a robot trained to play football using self-play simulation equivalent to 140 years of practice in weeks, illustrating a new approach to skill acquisition through computer simulation.

Skild AI trained a Unitree G1 robot to play soccer by simulating 140 years of practice against its own past versions, demonstrating significant progress in AI and robotics.

This paper introduces CoVer, a co-training framework for code generation that addresses self-play RL failures by using information-gain rewards and diversity-pruned tests, achieving significant pass rate improvements on benchmarks.

DiagEvo improves language-model self-evolution by deriving training direction from internal failure history via hierarchical error-cause memory and double-confidence filtering, outperforming baselines that rely on external resources.

This paper introduces ShuttleArena, a physics-based self-play environment for badminton where agents learn interpretable tactical policies using PPO, showing competitive improvement in shot selection and recovery.

LURE introduces a zero-data self-play framework for LLM reasoning that uses a pursuit-evasion game to dynamically adjust task difficulty and provide dense rewards, outperforming baselines across multiple environments.