PsyBer-Agent simulates user fatigue and preference drift to benchmark recommenders
UoLiverpool · hf · 2026-10-14
PsyBer-Agent (XJTLU, University of Liverpool, Alibaba) is a psychology-driven user simulator for recommender systems: its Psy-Engine models fatigue, affect and shifting preferences via three evolving latent states, calibrated against real logs through Unbalanced Gromov-Wasserstein optimal transport without fine-tuning the backbone LLM. It also releases the WebSim platform and PsyBer Benchmark (multimodal scenarios, seven recommender backbones), showing lower behavioral discrepancy than prompt-only simulators while sustaining up to 5M queued simulation tasks.
More from Research
- RUC and Kling AI propose EventLens: event-structure RL to fix video LLMs' temporal understanding — Guowen Zhang · 2026-10-14
- JSIR dataset drops: 267 jazz recordings with beat, chord and structure annotations — umpedronosapato · 2026-10-14
- There is no such thing as an AI drug, argues researcher Patrick Kidger — PatrickKidger · 2026-10-14
- More thinking, worse scores: refusals tank closed-model evals at max reasoning effort — _lewtun · 2026-10-14
- Integer multiplication repo becomes an agent harness benchmark — and collaboration loses — iskander · 2026-10-14
- Recurrent self-conditioning: framing recent diffusion methods as depth decoupled from time — alec_helbling · 2026-10-14