The non-obvious shift here is treating the agent harness as a live, mutable component rather than a fixed artifact frozen before deployment. Teams running multi-hour agent workloads — long coding jobs, research pipelines, infrastructure automation — can apply this pattern to let the agent adapt its own scaffolding without losing session state. It is most relevant to agent framework builders who currently restart or re-prompt when harness limits surface mid-run.
#长时程任务
9 stories · 1 in 7 days · 3 in 30 days
Stories, products and papers tagged "长时程任务", newest first.
Nov 10
Microsoft BOXBOX lets agents hot-reload their own harness mid-run
Oct 16
Chinese content论文显示:长任务中模型准确率的小差距会累积成巨大结果差异
Mollick 转了篇论文:长任务里便宜模型的错误会滚雪球,强模型当编排器才划算,做 agent 工作流的值得看看。
Oct 13
Chinese contentSambaNova 联合四所高校研究长时程任务下 LLM 智能体的有效性
SambaNova 拉上 Stanford、CMU 等四所高校发了篇论文,专门研究智能体跑长任务为什么会掉链子,做 agent 的可以看看。
Sep 21
Real-World Subtask RL for Long-Horizon Manipulation
This research improves robot task success rates by focusing reinforcement learning on critical subtasks, which could be relevant for anyone working on or interested in robotics and AI.
Sep 1
Google Agent Skills Paper
Available on arXiv, this method reduces token usage in long-running agents, making it practical for developers working with complex AI systems.
Agent Arena benchmarks agentic tasks
Developers can compare their models against others on practical, complex tasks through this benchmark.
Aug 26
Chinese contentRecursive Experiential-Working Memory Evolution for Long-Horizon Agent Harnesses
Recuris模型在长时程任务中显著提升了GPT-5.6 Sol和Claude Opus 5的表现,是长时程智能体驱动的递归记忆架构的一个突破。
Aug 18
Aug 5
Chinese contentLongHorizon-Harness:评测真实世界长时程智能体
有人发了 LongHorizon-Harness,一个专测长时程智能体的框架,论文在 Hugging Face,想评估多步任务表现可以看。