{"ID":2887817,"CreatedAt":"2026-06-01T04:54:23.091178241Z","UpdatedAt":"2026-06-01T04:54:23.091178241Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2508.00360","arxiv_id":"2508.00360","title":"Lucy: edgerunning agentic web search on mobile with machine generated task vectors","abstract":"Small language models (SLMs) are inherently limited in knowledge-intensive tasks due to their constrained capacity. While test-time computation offers a path to enhanced performance, most approaches treat reasoning as a fixed or heuristic process. In this work, we propose a new paradigm: viewing the model's internal reasoning, delimited by \u003cthink\u003e and \u003c/think\u003e tags, as a dynamic task vector machine. Rather than treating the content inside these tags as a mere trace of thought, we interpret the generation process itself as a mechanism through which the model \\textbf{constructs and refines its own task vectors} on the fly. We developed a method to optimize this dynamic task vector machine through RLVR and successfully trained an agentic web-search model. We present Lucy, a 1.7B-parameter SLM that leverages this dynamic reasoning mechanism with MCP integration to achieve 78.3% accuracy on the SimpleQA benchmark, performing on par with much larger models such as DeepSeek-V3. This demonstrates that small models can rival large ones when equipped with structured, self-constructed task reasoning.","short_abstract":"Small language models (SLMs) are inherently limited in knowledge-intensive tasks due to their constrained capacity. While test-time computation offers a path to enhanced performance, most approaches treat reasoning as a fixed or heuristic process. In this work, we propose a new paradigm: viewing the model's internal re...","url_abs":"https://arxiv.org/abs/2508.00360","url_pdf":"https://arxiv.org/pdf/2508.00360v1","authors":"[\"Alan Dao\",\"Dinh Bach Vu\",\"Alex Nguyen\",\"Norapat Buppodom\"]","published":"2025-08-01T06:45:29Z","proceeding":"cs.CL","tasks":"[\"cs.CL\"]","methods":"[\"Language Model\"]","has_code":false}