5 papers
Vernier: Probing Representational Misalignment Behind Lexical Gaps in Causal Reasoning
Zhenyu Yu
Instruction-tuned language models can answer the same causal-reasoning question differently after its English variable names are replaced by type-preserving placeholders, although…
InstantForget: Update-Free Backdoor Unlearning with Inference-Time Feature Reset
Zhenyu Yu
Backdoor unlearning aims to remove a malicious trigger behavior from a deployed model while preserving clean utility. We study the update-free inference-time setting, where model p…
When Does Latent Reasoning Help? MeRa: Metric-Space Bias for Spatial Prediction
Zhenyu Yu, Shuigeng Zhou
Latent reasoning has improved sequential recommendation by iteratively refining representations before prediction, but does it help spatial prediction? We find that the answer depe…
MARS: Multi-rate Aggregation of Recency Signals for Sequential Recommendation across Sparse and Dense Regimes
Zhenyu Yu, Shuigeng Zhou
Sequential recommenders weight historical interactions either through positional self-attention as in Transformers or through a single implicit decay schedule as in State-Space Mod…
Ghost: Plausible Yet Unlearnable Trajectories via On-Manifold Substitution for Next-POI Privacy
Zhenyu Yu, Jihong Guan, Shuigeng Zhou
A publisher who releases check-in trajectories inadvertently publishes a strong predictor of every user's future locations. We address this risk by generating unlearnable trajector…