SAGA: Scene-Aware, Goal-Evolving Agents for Long-Horizon CivRealm Strategy Planning

Long-horizon strategic planning in complex strategy games demands concurrent reasoning across multiple decision domains under imperfect information and sparse reward. Existing LLM-based agents suffer from three systematic failures: scene blindness from raw tile coordinates, context overflow and domain coupling from monolithic state dumps, and shallow cross-game learning that treats each episode in isolation. We present SAGA, an LLM multi-agent framework with three mechanisms each directly targeting one class of failure: (i) a Map-Semantic Scene Graph that encodes typed spatial relations among game entities into per-unit natural-language context, resolving spatial blindness without global token inflation; (ii) a Tool-Augmented Planner that pulls fine-grained domain state on demand and dispatches per-domain directives to dedicated specialist controllers, eliminating context overflow, domain coupling, and mechanical constraint violations; and (iii) a Dual-Horizon Feedback Loop that combines periodic within-game goal generation with structured cross-game causal post-mortem, enabling principled strategic evolution without manual reward engineering. Evaluated on FreeCiv, SAGA attains the highest mean civilization score -- the environment's sole sparse objective reward -- with lower variance than the two strongest baselines, and is the only method that significantly surpasses every baseline on infrastructure construction, the resource axis most readily sacrificed under multi-objective conflict. It outscores the two strongest baselines in most head-to-head games while cutting output tokens (the dominant decoding cost) by 27%. Equipped with the cross-game evolution module, SAGA reaches the highest end-of-chain score across five successive episodes. Ablation studies confirm that each architectural component contributes independently to this advantage.

Beyond tokens: a unified framework for latent communication in LLM-based multi-agent systems

Multi-agent systems built on large language models (LLMs) have become a prevailing paradigm for tackling complex reasoning, planning, and tool-use tasks. The dominant communication protocol in such systems is natural language: agents exchange messages token-by-token, verbalising their internal reasoning so that peers can read, verify, and respond. While convenient and interpretable, this protocol suffers from three structural drawbacks -- high inference cost, irreversible information loss during discretization, and ambiguity/redundancy of natural language. A growing body of work therefore explores an alternative protocol -- latent communication -- in which agents exchange continuous representations (embeddings, hidden states, or KV-caches) directly, bypassing the bottleneck of text generation. This paper presents a unified framework for organising the rapidly expanding literature on latent communication. We analyse existing methods along three orthogonal axes: (1) WHAT information is communicated (Embeddings, Hidden States, KV-Caches, or other continuous state); (2) WHICH sender-receiver alignment is used (latent-space alignment and layer alignment); and (3) HOW the communicated information is fused into the receiver (concatenation, prepending, mathematical operations, cross-attention, or cache restoration). Under this 3-axis framework, we systematically categorise eighteen representative methods proposed between 2024 and 2026, identify five major design patterns, and surface a set of open challenges -- including cross-architecture alignment, security of latent channels, compression for edge deployment, and the relationship between latent communication and latent chain-of-thought. We hope that this framework both lowers the barrier to entry for new researchers and provides a vocabulary for comparing future work.

RL-MSA: a Reinforcement Learning-based Multi-line bus Scheduling Approach

Multiple Line Bus Scheduling Problem (MLBSP) is vital to save operational cost of bus company and guarantee service quality for passengers. Existing approaches typically generate a bus scheduling scheme in an offline manner and then schedule buses according to the scheme. In practice, uncertain events such as traffic congestion occur frequently, which may make the pre-determined bus scheduling scheme infeasible. In this paper, MLBSP is modeled as a Markov Decision Process (MDP). A Reinforcement Learning-based Multi-line bus Scheduling Approach (RL-MSA) is proposed for bus scheduling at both the offline and online phases. At the offline phase, deadhead decision is integrated into bus selection decision for the first time to simplify the learning problem. At the online phase, deadhead decision is made through a time window mechanism based on the policy learned at the offline phase. We develop several new and useful state features including the features for control points, bus lines and buses. A bus priority screening mechanism is invented to construct bus-related features. Considering the interests of both the bus company and passengers, a reward function combining the final reward and the step-wise reward is devised. Experiments at the offline phase demonstrate that the number of buses used of RL-MSA is decreased compared with offline optimization approaches. At the online phase, RL-MSA can cover all departure times in a timetable (i.e., service quality) without increasing the number of buses used (i.e., operational cost).