Metraipolis › archive › dc_ea2b2b1d9832

Why Generate a Click? Plug-in Decision Heads for Web Agents — arXiv:2610.20107

Source
arXiv (en)
Original
https://arxiv.org/abs/2610.20107
License (as archived)
CC0-1.0
Archived
2026-10-15T20:35:28.006Z
SHA-256 of the archived text
343de74963637cbaa2e3b334dda9c04f7b96a5cec4f954572feaca75324fd64f
Stable URL of this page
https://metraipolis.com/documento/dc_ea2b2b1d9832

How to cite

Why Generate a Click? Plug-in Decision Heads for Web Agents — arXiv:2610.20107 — arXiv (en), https://arxiv.org/abs/2610.20107 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_ea2b2b1d9832 (Metraipolis archive dc_ea2b2b1d9832, sha256 343de74963637cbaa2e3b334dda9c04f7b96a5cec4f954572feaca75324fd64f)
BibTeX
@misc{dcea2b2b1d9832,
  title = {Why Generate a Click? Plug-in Decision Heads for Web Agents — arXiv:2610.20107},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.20107},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_ea2b2b1d9832; sha256 343de74963637cbaa2e3b334dda9c04f7b96a5cec4f954572feaca75324fd64f},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.20107. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Web agents built on vision language models decide every action by generating text such as click(12), although most of that text encodes a choice among a few action types and a few dozen marked elements. We read these choices directly from the hidden states that the generator computes while it reads the goal and the screenshot. Plug-in decision heads on a frozen generator predict the action type, pool the visual tokens inside each numbered box to pick the target, and point at a listed text option or at a WRITE slot that leaves the writing of new text to the generator. A confidence cascade executes the heads' action when they are sure and otherwise lets the generator decode from the KV cache of the same forward pass. With Qwen3-VL 4B and 8B generators trained on 213k web steps, a strict threshold keeps the cascade between 0.3 points below and 1.4 points above the generator on seven offline test sets while the heads decide 12 to 64% of the steps. A looser threshold raises step success on unseen Mind2Web websites by 3.1 and 1.8 points, costs up to 2 points on GUIAct and MiniWoB++, and lifts WebArena task success from 12.5 to 13.8% (4B) and from 13.6 to 14.6% (8B) with less than half of the generated tokens. Training the heads jointly with the generator lowers the generator's own step success by up to 4.9 points; heads on a frozen generator avoid this cost.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.CL, nuovi lavori del giorno).
Categorie arXiv: Computation and Language (cs.CL).
Pagina del lavoro: https://arxiv.org/abs/2610.20107 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-15T19:36:18.243Z dalla pagina https://arxiv.org/list/cs.CL/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).