New research from Meta. Agent harnesses are still mostly authored by hand. This makes it hard to tune robust agent harnesses for long-horizo…

New research from Meta. Agent harnesses are still mostly authored by hand. This makes it hard to tune robust agent harnesses for long-horizon tasks. In this new work, agents learn harness policies offline and deploy them to construct and update external harness state online during runtime task execution. EvoHarness-RL learns that policy instead. Belief, Progress, and Experience are exposed as harness state the policy can act on. Supervised harness fine-tuning teaches the action space, then cost-aware GRPO explores when to read, update, and consolidate during a long run. Qwen3-8B reaches 96.9% on ALFWorld. Two dynamics come out of the training. > Harness annealing means recurring harness-use patterns get absorbed into the model policy, and the agent shifts from frequent calls toward selective access. > Harness evolution means progress updates and experience consolidation compress the workspace into a compact task-adaptive state. This shows that long-horizon agents get more from a trainable coordination policy than from bigger tools or larger memories. Paper: https://arxiv.org/abs/2608.05446 Track more trending AI papers in our academy: https://academy.dair.ai/

Related

• STAPO: Selective Trajectory-Aware Policy Optimization for LLM Agent Training
• Very cool paper from Microsoft. The idea is to train agents on replayed teacher trajectories instead of live environment rollouts. On-policy…
• Policy-Conditioned Counterfactual Credit for Verifiable Reinforcement Learning of Long-Horizon Language Agents
• // State-Externalizing Harnesses // A new paradigm is emerging on how to effectively build agents and harnesses. If there is a state that th…
• Harness-1: Reinforcement Learning for Search Agents with State-Externalizing Harnesses

Source: DAIR.AI (X) | 2026-08-09