The authors tested how language agents (large language models) use interaction history within a single task.
Language agents often fail to use action outcomes from their history
Labeling observations as outcomes of prior actions improves task success and reduces repetition.
Chinese Tech
Jingyu Liu · Zhiwen Wang · Yuxin Jing · Huanyu Zhou · Yong Liu
Renmin University of China · ByteDance
Research Digest··2 min read
Liu et al.
Why this paper
From ByteDance and Renmin University of China · Part of Memory Management for Agents, now 49 papers
In one line
Language agents underuse action-outcome history; outcome labeling and a learned calibrator improve task success.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ·No stated limitations found
- ·No benchmark numbers found
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§