这篇研究很酷，RLM harness让Transformer在短任务上训练就能自动泛化到长任务，甚至跨领域。不用改模型结构，靠框架就能实现。
研究提出RLM harness作为组合泛化器，使Transformer无需额外泛化能力即可迁移任务。训练RLM时，短任务（如8倍长度）与长任务产生相同轨迹，模型可泛化至8-32倍长度未见任务。跨领域任务（如数学解法与论文写作）共享分解策略时也表现出相同泛化效果。该发现基于2026年论文，通过设计良好的harness形成商集，使LLM调用将结构相似任务视为相同。
Highly recommended. I've often claimed there's huge alpha in building agent harnesses. Turns out ...
Highly recommended. I've often claimed there's huge alpha in building agent harnesses. Turns out harnesses are compositional generalizers. The RLM harness is an instance of this. This could lead to interesting and efficient new ways to scale generalization in models. alex zhang @a1zhang Transformers struggle to generalize to tasks they were not explicitly trained on. Instead, we propose in 2026 that it is the job of the harness to generalize through composition. We observe a powerful property when training RLMs: for tasks with shared structure that look different, the root model naturally learns the same trajectory, meaning it views the two task trajectories as the same! In other words, the Transformer does not need additional generalization capabilities to transfer capabilities from one task to the other, the harness induces it. We find that well-designed harnesses form a quotient set over task trajectories, meaning their individual LLM calls can see structurally “similar” tasks as near-identical, token-for-token! Harnesses can effectively generalize for the Transformer during training, without relying on any intrinsic generalization capability from the model. For example, RLMs can see problems of different lengths as the same: we show that RLMs can train exclusively on short tasks, and fully generalize to similar but unseen tasks 8-32x longer because it produces near identical trajectories for both. Taking this further, we show that tasks across different domains (e.g. math solutions vs. essay writing) that share a decomposition strategy exhibit the same generalization effect. RLMs can train on the problem of finding which essays belong to the same author and improve performance on finding math problems that share similar solutions. The full blogpost, experiments, and discussion are in the thread below. 🔗 View Quoted Tweet 💬 0 🔄 0 ❤️ 0 ⚡