arXiv:2608. 03420v1 Announce Type: new Abstract: Large language models have improved substantially on single-shot reasoning tasks, but their performance in sequential decision-making is less well understood.
arXiv:2608. 12626v1 Announce Type: cross Abstract: Strategic reasoning in Large Language Models (LLMs) within long-horizon environments is often limited by inconsistent subgoals.
The paper introduces a text-based, multi-agent version of the board game Clue to test multi-step deductive reasoning in large language models (LLMs). Six LLM-based agents (GPT‑4o‑mini and Gemini‑2.5‑Flash) play turn‑based games, and a tool‑augmented approach uses a structured possibility matrix to convert implicit game state into explicit remaining possibilities, thereby offloading memory and deductive constraints from the agents. The study compares this tool‑augmented method against a baseline to assess its impact on reasoning quality and task success in a strategic reasoning environment.
The paper introduces the Abstraction Agent, a zero‑shot pipeline that employs a large language model to automatically generate continuous strategic features from a natural‑language game description, score private states, and cluster them into abstraction buckets without any game‑specific evaluators or training data. The pipeline consists of four phases—feature discovery with calibration anchors, batched private‑state scoring, correlation‑based feature selection, and k‑means clustering—and achieves significant reductions in lifted‑strategy exploitability in heads‑up no‑limit Texas hold’em and outperforms scalar rank baselines in ROVER Trials. The method also transfers to other games such as four‑card Pot‑Limit Omaha, HUNL preflop and flop, and Riichi Mahjong, demonstrating that it can uncover strategic concepts that align with recognized game theory insights.
arXiv:2607. 25308v1 Announce Type: cross Abstract: Training large language models (LLMs) to act in long-horizon games is a promising step toward generalist decision-making, yet reinforcement learning with verifiable rewards (RLVR) relies on sparse final rewards that reveal little about which decisions determine success.
arXiv:2609.38881v1 Announce Type: new Abstract: Real-time strategy (RTS) games require agents to coordinate economic development, production and construction, base defense, unit organization, and att...
arXiv:2605.09278v2 Announce Type: replace Abstract: Multi-agent debate (MAD) systems increasingly rely on shared memory to support long-horizon reasoning, but this convenience opens a critical vulner...
arXiv:2608. 06741v1 Announce Type: new Abstract: Reasoning in large language models (LLMs) is often grounded in human text, human demonstrations, and human-generated rationales.
Self-Play Search Distillation (SPSD) is a framework that generates superhuman synthetic data by having MuZero-like networks play board games in executable environments. The search records are converted into structured reasoning chains that serve as environment‑grounded supervision for training large language models. When applied to Qwen3‑4B‑Base, SPSD improves performance on six mathematics benchmarks from 24.1 to 36.6 and raises the win rate on unseen games from 15% to 45%.
arXiv:2606. 00017v1 Announce Type: new Abstract: Training language model agents for multi-agent strategic interaction presents a core difficulty: the quality of any action may depend on future events that never materialize, on moves that violate game rules, or on decisions made by other players.
arXiv:2608.21871v1 Announce Type: new Abstract: Reinforcement learning with verifiable rewards has become the dominant recipe for improving large language model reasoning, yet it presumes large human...
ArenaFlow is a hierarchical credit propagation framework designed to improve reinforcement learning for open-ended agent tasks. It uses tournament-based relative ranking to generate trajectory-level rewards and structured reflective evaluation to identify pivotal success steps, reusable strategy skills, and skill usage attribution. The framework propagates advantages to high-confidence steps and maintains a global skill memory, enabling more targeted optimization and reusable skill priors for future exploration.
arXiv:2510. 05592v2 Announce Type: replace Abstract: Outcome-driven reinforcement learning has advanced reasoning in large language models (LLMs), but prevailing tool-augmented approaches train a single, monolithic policy that interleaves thoughts and tool calls under full context; this scales poorly with long horizons and diverse tools and generalizes weakly to new scenarios.