AI
SimTrace: Grounded Multimodal User Trajectories Generation for Online User Modeling
Publication: cs.AI updates on arXiv.org
Author: Yunan Lu, Shuang Xie, Meghna Allamudi, Mingyu Zhao, Han Li, Lingyun Wang, Zhou Yu
Source published: 01/10/2026, 04:00:00 UTC
Briefing created: 01/10/2026, 04:05:05 UTC
Based on a feed excerpt; full article content has not been acquired. Source evidence retrieved 01/10/2026, 04:03:25 UTC.
Briefings
SimTrace is an open-source framework that generates fine-grained synthetic multimodal clickstreams using a grounded computer-use agent. It addresses the scarcity of real user data by anonymizing interactions and simulating a web environment twin. Evaluated in an e-commerce setting, SimTrace outperformed baseline models on 7 of 8 fidelity metrics and improved next action prediction accuracy by 11.0% when augmented with real data.
Key points
- SimTrace generates synthetic clickstreams by pairing anonymized real interactions with a simulated web environment twin.
- Models trained on SimTrace data achieved performance comparable to real data on purchase prediction and recommendation tasks.
Context and limitations
Evaluation was limited to an e-commerce setting; results may not generalize to other domains or platforms.
Generated by Bittrees-hosted qwen35-2b sha256:aaf42c8b7c3c on 01/10/2026, 04:05:05 UTC. Based on publisher-provided feed evidence; may not cover the full original article. Check the source for details.
View archived briefing on IPFS
bafybeiekxo6fow5bdrxldyhtnezjvqp53evbujdgia6445rpyjfbfdkeei