AI Science Tracker
EN

WiSI: Wireless-in-the-Loop Self-Improvement

Department of Electrical and Computer Engineering, The University of Hong Kong · 3 Nov 2026 · computer science

Discovery

WiSI: Wireless-in-the-Loop Self-Improvement

Source check: Source checked. Not independent scientific confirmation.

Finding: WiSI evaluates language-model-driven program improvement under wireless execution feedback. In the reported tasks, agents improved under ideal links achieved 40.0% success under congestion, compared with 77.2% for agents improved under realistic wireless conditions. The result is limited to the tested setup.

Technology: Large language model-driven program improvement evaluated in a wireless-in-the-loop framework

Publication: Preprint · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
WiSI uses an LLM to propose program changes; execution feedback is generated under modeled wireless conditions and informs candidate selection.
Test and result
A controlled comparison found 40.0% task success under congestion for agents improved under ideal links, versus 77.2% for agents improved with realistic wireless conditions. The authors report retained communication-efficient program structures.
Limitation / still unknown
Results are limited to the studied tasks, agent and wireless simulation; real-world deployment generalization is not established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/abs/2611.04437v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 3/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Open in the full catalog for filters, charts and related discoveries. Browse computer science discoveries.