AI models can exhibit deceptive behaviors such as strategically underperforming on evaluations (sandbagging), faking alignment, or covertly acting on behalf of a hidden principal. Safe deployment therefore requires auditing techniques that can reliably detect such behaviors. To develop and test these techniques, we can...
Vasisht Duddu, Mark Wang, Lipeng He et al.· 0 citations
In latent generative models, reconstruction quality is often assumed to correlate with generative performance. However, reconstruction FID (rFID) can exhibit weak or even negative correlation with generation FID (gFID). We attribute this misalignment to a latent distribution mismatch: reconstruction evaluates the decod...
Xianghong Fang, Wenjie Shu, Tong Xu et al.· 1 citation
Latent audio generative models are typically trained in two stages: an audio codec is learned first, followed by a latent generative model. This decomposition leads to a decoder train-generation mismatch: the codec decoder is trained on encoder-induced latents but deployed on generator-produced latents at inference tim...
Xianghong Fang, Geeyang Tay, Wen-Tao Ma et al.· 0 citations
This work proposes MedCertAIn, a predictive uncertainty framework that leverages multimodal clinical data to improve model performance and reliability for in-hospital mortality risk prediction as an indicator of patient deterioration and highlights the promise of data-driven priors in advancing robust, uncertainty-awar...
L. J. L. López, Tim G. J. Rudner, Farah E. Shamout· 0 citations
By repurposing Selective Gradient Masking, a suite of six 254M-parameter language models on English Wikipedia with graded levels of disentanglement between biology and non-biology knowledge are trained, it is found that more disentangled models consistently achieve better retain-forget trade-offs.
Evžen Wybitul, Tim G. J. Rudner, C. D. de Witt· 0 citations
Discrete visual tokenization, predominantly driven by vector, scalar, and product quantization, lacks a unified conceptual framework for understanding quantization tradeoffs. In this paper, we propose a unified rate--distortion perspective on modern discrete visual tokenization. By viewing quantization as lossy compres...
Xianghong Fang, Wenlong Mou, Yuan Yuan et al.· 0 citations
VQ-Transplant democratizes quantization research by enabling resource-efficient integration of novel VQ techniques while matching industry-level reconstruction performance.
Xianghong Fang, Yuan Yuan, Dehan Kong et al.· arXiv.org· 1 citation
Relay introduces a differentiable per-token channel that passes information between forward passes and is trained via truncated backpropagation through time (BPTT), demonstrating that state-of-the-art DLMs can be explicitly trained to relay latent information forward across decoding steps, advancing the performance-lat...
Benjamin Rozonoyer, Jacopo Minniti, Dhruvesh Patel et al.· arXiv.org· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.