Gan et al.
Web agents can skip text generation and read action choices from hidden states
Plug-in decision heads recover action type, target element and option from the prefill of a frozen vision language model, with decoding reserved for uncertain steps and open text.
Research Lab
Chengguang Gan · Yunhao Liang · Hanjun Wei · Qinghao Zhang · Zhixi Cai · Shiwen Ni
Independent Researcher · University of Chinese Academy of Sciences · Pusan National University · Monash University · Shenzhen University of Advanced Technology
Research Digest··3 min read
Gan et al.
Why this paper
From University of Chinese Academy of Sciences and 4 others
In one line
Plug-in decision heads read web agent actions from hidden states, enabling faster decisions on 12-64% of steps without loss of accuracy.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ✓Compute or model size stated (params 4B)
- ✓Limitations stated by the authors
- ✓Reports numbers on named benchmarks (2 benchmarks)
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§