Search papers, labs, and topics across Lattice.
This paper introduces Agent-as-Annotators, a structured synthetic trajectory generation framework for web agents that decomposes the process into modular LLM components analogous to human annotation roles. They use Gemini 3 Pro to generate 3,000 trajectories across six web environments and fine-tune a 9B-parameter student model via supervised learning on high-quality trajectories. The resulting student model achieves 41.5% on WebArena, surpassing Claude 3.5 Sonnet and GPT-4o, and demonstrates strong generalization to unseen environments like WorkArena L1.
A 9B parameter model, distilled from a single frontier LLM using structured synthetic trajectories, outperforms both Claude 3.5 Sonnet and GPT-4o on web navigation tasks.
Frontier LLMs can navigate complex websites, but their cost and reliance on third-party APIs make local deployment impractical. We introduce Agent-as-Annotators, a framework that structures synthetic trajectory generation for web agents by analogy to human annotation roles, replacing the Task Designer, Annotator, and Supervisor with modular LLM components. Using Gemini 3 Pro as teacher, we generate 3,000 trajectories across six web environments and fine-tune a 9B-parameter student with pure supervised learning on the 2,322 that pass quality filtering. The resulting model achieves 41.5% on WebArena, surpassing closed-source models such as Claude 3.5 Sonnet (36.0%) and GPT-4o (31.5%) under the same evaluation protocol, and nearly doubling the previous best open-weight result (Go-Browse, 21.7%). Capabilities transfer to unseen environments, with an 18.2 percentage point gain on WorkArena L1 (an enterprise platform never seen during training) and consistent improvements across three additional benchmarks. Ablations confirm that each pipeline component contributes meaningfully, with Judge filtering, evaluation hints, and reasoning traces each accounting for measurable gains. These results demonstrate that structured trajectory synthesis from a single frontier teacher is sufficient to produce competitive, locally deployable web agents. Project page: https://agent-as-annotators.github.io
