arXiv:2607. 08964v1 Announce Type: new Abstract: AI agents have become capable of autonomously completing short, well-specified tasks.
arXiv:2606. 07682v1 Announce Type: cross Abstract: AI agents are increasingly expected to complete long-horizon workflows that require sustained progress over hours, millions of tokens, and complex environments.
arXiv:2610.18868v1 Announce Type: cross Abstract: Long-horizon agents have drawn growing attention for their ability to work autonomously -- from hours to days -- on economically valuable tasks, a ca...
The paper introduces a self‑evolving harness framework where a frozen language‑model agent first solves tasks and then edits its own harness based on run records. Using a 49‑line seed harness, the evolved harness improves average scores on in‑distribution benchmarks by 4.48 points and on out‑of‑distribution benchmarks by 12.64 points, surpassing Codex on the former and matching it on the latter. Continued evolution on a specific out‑of‑distribution benchmark further raises performance, and the study analyzes emergent mechanisms such as output truncation and history compaction.
Long-horizon agents have drawn growing attention for their ability to work autonomously -- from hours to days -- on economically valuable tasks, a capability essential to automated discovery and recur...
arXiv:2608. 16798v1 Announce Type: cross Abstract: Agent harnesses have substantially improved performance on long-horizon tasks by coordinating agent interactions with the environment.
The paper introduces PARTS, a real‑world subtask reinforcement learning framework that fine‑tunes a pretrained robot policy by focusing on critical bottleneck subtasks while keeping the base policy frozen. It uses agent‑generated selectors and success verifiers to provide local rewards, enabling learning even when full‑task successes are rare. Experiments on bimanual YAM and single‑arm Franka robots show that PARTS raises complete‑task success from 32% to 61% and from 50% to 95%, respectively, with only tens of minutes of real‑world RL rollouts and minimal human intervention.
OS-Marathon is a new benchmark that tests computer‑use agents on vast‑horizon, repetitive tasks, covering 100 tasks across five scenarios and ten domains. The study shows that current state‑of‑the‑art agents perform poorly on these tasks, and that simply decomposing workflows into subtasks does not solve the problem. Introducing a cost‑friendly personalization method called GraphDemo, which adapts agents from a single human demonstration, improves performance, highlighting the value of human guidance for these challenging tasks.
arXiv:2607. 05377v1 Announce Type: cross Abstract: While recent Vision-Language-Action (VLA) models show promise toward generalist manipulation policies, they struggle with long-horizon tasks due to their Markovian nature-relying solely on current observations.
arXiv:2605. 26494v2 Announce Type: replace-cross Abstract: We introduce the MiniMax-M2 series, a family of Mixture-of-Experts language models built around the principle that mini activations can unleash maximum real-world intelligence.
The paper introduces T1, a 122‑billion‑parameter Mixture‑of‑Experts model trained with reinforcement learning to perform long‑horizon terminal tasks such as coding and scientific discovery. T1 operates a real shell in a cloud sandbox, making over 300 tool‑call turns per task and receiving rewards from task‑specific verifiers. The authors detail a training recipe that includes aggressive warm‑starting, TITO construction with drift repair, and rollout‑routing replay, achieving significant performance gains on Terminal‑Bench 2.1 and surpassing GPT‑5.4 and GLM‑5.1 on the Long‑Horizon Terminal Bench.
TrajEvolve introduces a trajectory evolution paradigm with hindsight credit assignment to improve reinforcement learning for large language models on long-horizon tasks. It refines low-quality steps in historical trajectories to expose the policy to high-quality compositions that are rarely sampled, and reformulates step-level continuous credit assignment as a binary classification problem to identify redundant steps. Experiments on two datasets show that this approach outperforms existing methods.