ScholarPulse 日报 2026-10-14

2026-10-14 学术简报：2 篇。ChunkAct 将动作分块引入智能体机器人控制，通过离线将记忆编译为动作块并由在线自主执行，大幅降低大语言模型计算开销并提升任务成功率。

今日速览

序号	标题	来源	日期	主题	推荐等级
1	ChunkAct: Action Chunking for Agentic Robotics	arXiv	2026-10-13	AI-Agent	高
2	Hybrid Reasoning Agents	arXiv	2026-10-13	AI-Agent	高

重点论文与技术动态

1. ChunkAct: Action Chunking for Agentic Robotics

来源：arXiv
日期：2026-10-13
作者/机构：Bingxin Xu, Wenrui Bao, Xinxin Liu, Jiawei Zhou, Yuzhang Shang, Emilio Ferrara
主题标签：AI-Agent, arXiv
推荐等级：高
分类：cs.RO, cs.AI, cs.CL, cs.CV

一句话结论

ChunkAct 将动作分块引入智能体机器人控制，通过离线将记忆编译为动作块并由在线自主执行，大幅降低大语言模型计算开销并提升任务成功率。

核心内容

消除推理冗余：克服每个动作基元均需 LLM 遍历交互历史推理的高成本弊端，使计算开销从基元数量级降至决策次数级。
动作分块架构：离线将记忆编译为动作块以实现测试期免记忆检索；在线单次调用即可自主执行链式基元，仅在失败时交还控制权。
高效且高成功率：在 LIBERO-Pro 基准上成功率较 SOTA 提升 5.0 点，同时大模型调用减少 2.9 倍，输入与输出 Token 分别减少 4.4 倍和 2.3 倍。

方法与数据

提出动作分块编译与在线链式基元自主执行机制；实验评测基于 LIBERO-Pro 基准。

价值判断

值得关注：解决了智能体机器人在动作层级频繁调用 LLM 带来的严重效率瓶颈，兼顾了泛化性与执行效率。
可复用点：将经验/记忆离线编译为动作块的设计，以及“单次触发、自主链式执行、仅失败回退”的控制交互范式。
局限/待核查：仅在仿真基准 LIBERO-Pro 上进行了验证，真机部署效果及复杂现实环境中的失败恢复机制在摘要未明确。

摘要

Agentic robotics places a large language model (LLM) agent in the robot control loop, where it composes perception, motion, and vision-language-action (VLA) tools under the guidance of a memory and thereby generalizes to new tasks without retraining. This generality comes at a steep price in efficiency. Every primitive requires a round of LLM reasoning over the full interaction history, so the cost of an episode grows with the number of primitives. We find this reasoning largely redundant: once a sub-goal is chosen, its primitives follow a routine already prescribed by memory. We present ChunkAct, which introduces action chunking to agentic robotics. Online, a single call launches a chunk of chained primitives, which executes autonomously and returns control only upon failure. Offline, ChunkAct compiles the memory into these chunks, so execution is memory-free at test time and the planner is reserved for decisions that require reasoning. LLM computation therefore scales with the number of decisions rather than the number of primitives. On LIBERO-Pro, ChunkAct improves task success by 5.0 points over the state of the art in agentic robotics while requiring 2.9x fewer LLM calls, 4.4x fewer input tokens, and 2.3x fewer output tokens.

2. Hybrid Reasoning Agents

来源：arXiv
日期：2026-10-13
作者/机构：Jiaru Zou, Yunzhe Qi, Sirui Chen, Federico Bianchi, Shang Zhu, Pan Lu
主题标签：AI-Agent, arXiv
推荐等级：高
分类：cs.CL, cs.AI, cs.LG

一句话结论

HybridAgents 通过在隐空间推理与显式文本交互间动态切换，构建了兼具原生工具调用能力、高透明度与高推理效率的混合智能体架构。

核心内容

架构协同：通过轻量级 HybridLink 和连续工作记忆，实现隐式推理与文本模式的无缝转换及跨轮次无损上下文保留。
能力兼顾：突破了传统隐空间推理的壁垒，在原生集成外部工具的同时，使隐空间思维过程具备直接可审查性。
显著增效：综合性能提升 10.4%，Token 消耗减少 35.5%~~54.7%，推理提速 1.8~~2.6 倍，成本降低 2.7~3.7 倍。

方法与数据

提出专用后训练算法 HybridSFT（稳定混合生成）与 HybridRL（优化思考与行动决策）；在 3 种模型规模及数学、医学等 11 个基准上完成评测，代码已开源。

价值判断

值得关注：打通了高效隐式推理与环境交互行动的边界，大幅降低智能体部署成本。
可复用点：HybridLink 结构设计、混合工作记忆机制与两阶段混合后训练算法。
局限/待核查：具体模型底座及隐空间思维如何“直接审查”的实现细节摘要未明确。

摘要

How can we build an agent that thinks beyond human language, yet still acts through it in the external environment? Modern LLM agents still reason and interact through text, while latent reasoning offers an efficient internal medium but lacks native tool access and reasoning transparency. We bridge these two directions with Hybrid Reasoning Agents (HybridAgents), which dynamically navigate between latent reasoning and text-based interaction throughout a unified generation process. Architecturally, a lightweight HybridLink enables seamless transitions between text and latent modes, while the agent's continuous working memory preserves lossless hybrid context across multiple turns. These designs preserve tool-integrated agentic capabilities while making latent thoughts directly inspectable. Building on this architecture, we further develop dedicated post-training algorithms for hybrid reasoning: HybridSFT stabilizes tool-integrated hybrid generation by teaching the backbone to coordinate with HybridLink, then HybridRL scales reasoning by learning when to think in latent space and when to act through text. We evaluate HybridAgent at 3 model scales across 11 benchmarks covering mathematics, science, medicine, search, and agentic reasoning. Compared with frontier LLM Agents and agentic frameworks spanning both text-based and latent methods, HybridAgent consistently improves performance by 10.4%, while delivering 35.5%-54.7% token reduction, 1.8x-2.6x faster inference, and 2.7x-3.7x cost savings. Code and data are available at https://github.com/togethercomputer/HybridAgent.