Jianing Zhu

Unveiling Representation Expansion: A New Lens on Out-of-Distribution Detection

Taming CoT Obfuscation in VLMs: From Mechanistic Evidence to Activation Enforcement

This work proposes Targeted Anti-obfuscation with Mechanistic Enforcement (TAME), which uses Sparse Autoencoders (SAEs) to combine behavioral feedback with targeted suppression of template-associated activations during RL, providing a path from behavioral monitoring to representation-level oversight for more auditable...

Towards Understanding On-Policy Distillation through the Lens of Test-Time Scaling

It is observed that OPD-trained models maintain superior avg@K performance across sampling budgets, while the advantage in pass@K gradually shifts to the pre-OPD base models as K increases, suggesting that OPD primarily improves sampling efficiency rather than consistently expanding the student's reasoning capability b...

POPS: Recovering Unlearned Multi-Modality Knowledge in MLLMs with Prompt-Optimized Parameter Shaking

This paper proposes a novel adversarial strategy, namely Prompt-Optimized Parameter Shaking (POPS), aiming to recover the supposedly unlearned multi-modality knowledge from the MLLMs, exposing fundamental vulnerabilities that challenge the foundational robustness of representative MMU-based privacy protections.