这篇论文用Qwen3-8B做了AppWorld上的合并vs联合训练对比，发现合并效果不输联合训练，而且解释了为什么不同合并方法结果差不多。
研究首次将模型合并与联合多任务训练在强化学习场景下直接对比。使用Qwen3-8B专才在AppWorld智能体基准上通过LOOP训练难度1和难度2任务，合并方法（TIES、RAM+）与联合训练模型相比，任务目标完成率无统计差异。任务向量余弦相似度仅0.06-0.10，近正交但支撑集重叠约65%，导致支持&符号合并退化为近均匀平均。所有代码和统计数据已开源。
When Model Merging Rivals Joint Multi-Task Reinforcement Learning: A Task-Vector Geometry Analysis
Model merging is promoted as a substitute for joint multi-task training, yet in the reinforcement-learning setting this substitution is essentially never tested against the baseline it claims to replace: methods merge independently released agents precisely because a joint model is unavailable. We build the missing comparison. Training difficulty-1 and difficulty-2 Qwen3-8B specialists on the AppWorld agent benchmark with LOOP, we merge them (TIES, RAM+) and pit the result against a jointly trained model on the same data. On task-goal completion, merging matches joint RL -- and every merge variant is statistically indistinguishable. To explain why merge method does not matter here, we measure the geometry of the specialists' task vectors, which carries no task-sampling noise: they are near-orthogonal (cosine 0.06 - 0.10) despite ~65% support overlap, a small, shared direction that grows over training and that we calibrate against a random-init floor and a same-run ceiling to confirm it reflects learning, not the low-rank parameterization. Because direction and support are decoupled, support and sign-based merging (RAM, TIES) collapse to near-uniform averaging. We release all code and statistics.