This work proposes a method that converts complex agentic prompts into minimal contrastive pairs in which a single request verb determines the tool-call decision, and traces the decision to a vector, $\mu_\Delta$, that is both causally necessary and sufficient and generalizes beyond the discovery prompts to native mult...
X. Gong, Ting-Xu Han, Jia-Hao Zhang et al.· 0 citations
Sparse autoencoders (SAEs) reveal visual structure in diffusion transformers (DiTs), but interpreting a feature does not establish whether it can be used to control generation. We introduce D-Scope (Diffusion Scope), a framework that connects feature interpretation to generation control through shared visual evidence....
Xinyue Xu, Jiahao Zhang, Li-Jie Hu et al.· 1 citation
Interpretability methods such as probes, activation patches and learned editors are designed to reveal or modify a model's current computation. World models pose a harder requirement: because their predictions become inputs to later predictions, a useful internal correction must survive after editing stops. We therefor...
Jun-Chi Yao, Zi-Yi Wang, You-Ling Huang et al.· 0 citations
Linear probing and activation steering use linear directions to predict and control behavior-level concepts such as correctness, safety, and social bias in question answering. We call these \emph{behavior-level concept directions}. Drawing on the Linear Representation Hypothesis (LRH) and circuit studies, these directi...
Man-Jiang Yu, Hong-Ji Li, Zi-Han Wang et al.· 0 citations
Experiments show that CleanVideo effectively erases target concepts while maintaining visual fidelity and temporal coherence, outperforming existing baselines under frame-level and video-level evaluations and under concept-recovery attacks when the protected pipeline remains intact.
Jun-Hui Liao, Hongji Li, Wen-Rui Zhou et al.· 0 citations
Experiments on BUS-CoT and IU X-ray datasets demonstrate consistent improvements in diagnostic accuracy, concept consistency, and report quality over strong general-purpose and medical MLLMs, indicating that concept-grounded reasoning better aligns generation with clinical decision processes.
Xin-Yue Xu, Hong-Bin Lin, Juan-Gui Xu et al.· 0 citations
This work shows that a single shared SAE can replace a collection of dedicated per-model SAEs, and combines a shared dictionary with model-specific encoder-decoder pairs to achieve near-dedicated-SAE reconstruction quality.
Daniil A. Ognev, Célian Vasson, Li-Jie Hu et al.· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.