Decision level continual learning improves personalization from typed user feedback

Jev-DLC: Decision-Level Continual Learning from Typed Feedback

Artificial Intelligence

Summary

Personal preferences often only become clear after a system starts being used, when users accept or reject certain options. The paper presents a way to learn continually from this kind of user feedback, called Jev-DLC, which keeps track of corrections without storing all past data or doing heavy computations later. It uses the feedback to better predict what the user prefers in future decisions. The authors also created a benchmark to fairly measure improvements, showing Jev-DLC performs better than earlier methods, especially as more feedback accumulates.

What this means in practice

  • For software personalization teams: Enhance decision-making systems by using typed user corrections to continually adapt without extra runtime cost or large storage needs.
  • For digital assistant developers: Improve personalization of assistant responses over time by leveraging explicit feedback that directs preferred choices without rerunning complex models.

Authors

Yangyang Feng, Jianle Gan, Zhuoyan Feng, Chengwei Qin

Abstract

Individual preferences often become visible only after deployment, when a user accepts one decision or revokes it in favor of another. Jev-style typed interfaces make this feedback explicit by identifying the selected action and, under revocation, the decision it replaces. We study how these consequences persist across requests as Decision-Level Continual Learning from Typed Feedback (Jev-DLC). Jev-DLC treats each correction as both evidence about a recurring action identity and a directed relation between the preferred and rejected choices. It preserves these signals in complementary bounded channels above a frozen scorer and combines their scores to shape the next decision. The method retains no raw interaction history, adds no scorer call, and performs no runtime language-model backward pass. We also introduce TCPA-Bench, a chronological suite that credits a correction only through its effect on later decisions. On TCPA-Repeat, Jev-DLC reaches 44.83% user-macro accuracy overall, 2.43 points above Recent-4 history replay. After at least two earlier feedback events, it reaches 56.07%, a 7.54-point margin over the strongest reported comparison at this depth. These results show how explicit typed corrections can become compact causal state for continual personalization.