A Novel Task-Driven Method with Evolvable Interactive Agents Using Event Trees for Enhanced Emergency Decision Support

First Xingyu Xiao, Second Peng Chen, Third Ben Qi, Fourth Jingang Liang, Fifth Jiejuan Tong, Sixth Haitao Wang · 2025

Read A Novel Task-Driven Method with Evolvable Interactive Agents Using Event Trees for Enhanced Emergency Decision Support by First Xingyu Xiao, Second Peng Chen, Third Ben Qi, Fourth Jingang Liang, Fifth Jiejuan Tong, Sixth Haitao Wang (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2501-06193.txt.

Description

When a complex plant fails, operators must decide fast with incomplete news, and preset plans rarely fit the accident that actually happens. Xingyu Xiao and co-authors build a decision helper in a paper whose head record gives 2024. Two kinds of language-model agents share the work: doers that carry out steps and checkers that judge them, guided by event trees that map how accidents branch.

The agents learn from both their hits and their misses across three tasks, from splitting an event into parts to recommending a call. In tests on nuclear plant scenarios the system reached perfect scores on some unseen incidents and beat older approaches. The authors present the design as a way to keep human operators ahead of a crisis rather than buried by it.

Catalog record

CREATORFirst Xingyu Xiao, Second Peng Chen, Third Ben Qi, Fourth Jingang Liang, Fifth Jiejuan Tong, Sixth Haitao Wang
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE46 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2501-06193

Browse the full library · About the Tower of Babel