).
Outcome verification reveals failures in social agent plans that conversation ratings miss
A new benchmark and training method use paired scenarios with private conditions to assess and improve plan feasibility
Chinese Tech
Caixin Kang · Ke Wang · Yuchuan Wu · Xiaoqiang Wang · Ling Li · Bingxue Qiu · +2 more
Alibaba Group · The University of Tokyo
Research Digest··3 min read
Kang et al.
Why this paper
From Alibaba Group and The University of Tokyo
In one line
Social agents that sound successful often produce invalid plans, while paired-world outcome training improves verified coordination.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ✓Limitations stated by the authors
- ✓Reports numbers on named benchmarks
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§