1 paper · 1 filter
Udit Thakur, Baohao Liao, Matteo Nulli +4
Search-augmented language models increasingly serve as long-horizon research agents, yet most open training methods assume a fixed user goal throughout search. We study mid-respons…