Tag
提出多时间尺度分层世界-动作模型（MHWA），通过分层时间抽象缓解自回归世界模型开环预测中的漂移与误差累积问题，在样本高效决策上优于扁平基线。
The paper introduces Generalized Implicit Temporal Abstraction (GITA), a method for offline goal-conditioned reinforcement learning that trains a single k-conditioned value function across multiple temporal abstraction scales, resolving the trade-off between long-range value signal and local resolution. GITA outperforms offline GCRL baselines like HIQL and OTA on OGBench, raising average success rates by 25 percentage points over HIQL.
This paper analyzes the Option-Critic framework in reinforcement learning, explaining that the learned termination rule is redundant, introducing 'policy necrosis' where options fail to explore, and showing that extra options improve performance by reducing the probability of joint failure.
SPACE 通过从成功轨迹归纳两层参数化技能，将子技能边界作为动作块监督，训练长程 LLM Agent 自适应输出可变长原子动作序列；在 ALFWorld 和 ScienceWorld 上成功率提升 7.0%–31.3%，决策轮次最多降低 78.9%。
This paper proposes a two-level hierarchical reinforcement learning framework called ToSCA for conversational agents, which uses DQN and PPO with a dual-granularity reward mechanism to improve strategy determination and response quality in multi-turn conversations.
Rich Sutton discusses the common mistake of relying on one-step predictions in AI research, advocating for temporally abstract models using options and GVFs.