The Interplay of Harness Design and Post-Training in LLM Agents
For researchers building tool-integrated LLM agents, this work highlights the importance of considering harness design during post-training, especially for robustness to environment shifts.
The paper extends ALFWorld to treat harness design as a controllable dimension and evaluates post-training under task and tool environment shifts. It shows that harness-aware post-training improves in-distribution performance and enables robust adaptation to out-of-distribution settings, while minimal harness design leads to drastic performance drops under stronger shifts.
Tool-integrated LLM agents are often wrapped within a harness: the scaffolding that determines which tools are exposed, how they are described, and what auxiliary information accompanies each per-step observation. While agents are routinely post-trained, this scaffolding is typically treated as a fixed engineering detail, with design effort limited to the training-free regime. Moreover, existing post-training algorithms assume a static environment, even though tool environments and tasks often shift upon deployment. To address this gap, we extend $\texttt{ALFWorld}$ (i) to treat the harness as a controllable design dimension and (ii) to support evaluation under task and tool environment shifts. Building on this, we systematically analyze how the harness design influences post-training in both in-distribution and out-of-distribution (OOD) settings. We empirically show that harness-aware post-training not only improves in-distribution performance but also enables agents to robustly adapt to OOD settings. Under a harness with minimal design effort, post-training suffers a drastic performance drop under stronger tool environment shifts, further highlighting the importance of harness-aware post-training under such shifts.