The authors first tested whether earlier reasoning continues to support actions across multiple interaction turns.
Agents can skip repeated reasoning when earlier plans still suffice
RACE trains language-model agents to estimate how long prior reasoning remains useful, reducing repeated reasoning across four interactive benchmarks.
Chinese Tech
Yiruo Cheng · Shen Huang · Xiaoshuai Song · Jiejun Tan · Guanting Dong · Pengjun Xie · +2 more
Renmin University of China · Alibaba Token Hub, Alibaba Group
Research Digest··2 min read
Cheng et al.
Why this paper
From Alibaba Token Hub, Alibaba Group and Renmin University of China
In one line
Adaptive reasoning can be trained by estimating whether earlier reasoning still supports later actions, a likelihood signal that cuts reasoning cost while preserving performance.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ·No stated limitations found
- ✓Reports numbers on named benchmarks (4 benchmarks)
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§