Randomly initialised neural networks of width $m$ converge to Gaussian processes as $m \to \infty$, with two types of quantitative bounds characterising this convergence. Finite-sample bounds compare the network with its Gaussian limit at a fixed finite sample set, on an arbitrary domain, in the quadratic Wasserstein d...
Luis Wiederhold, Nil Ayday, Debarghya Ghoshdastidar· 0 citations
On-policy self-distillation improves a model using guidance from the model itself, but it still requires a teacher that provides more useful predictions than the student. Existing approaches obtain such guidance by giving the teacher additional context or instructions. We ask whether it can instead come from changing h...
Wei Fan, Qing-Yue Yang, Chao-Hong Tan et al.· 0 citations
Sycophancy in Large Language Models, the tendency to align responses with user preferences over ground truth, has largely been studied using questions without a clear ground truth and with a single, plainly stated user opinion. We present a systematic study of which framings of a user's claim trigger sycophancy in fact...
Jean Seo, Pepa Atanasova, Isabelle Augenstein· 0 citations
Owner-Aware Routing is proposed, which drafts a schema-only SQL query, counts the owners whose tables it joins, and routes the question before any private row is read, and beats every fixed protocol on PeopleJoin, by up to 9.3 points on held-out organizations.
Zhe-Xi Feng· 0 citations
Reach audiences
Advertise in front of researchers, engineers, and readers.
Standard identifiability theory assumes one-dimensional latent factors, but learned representations can encode latent information across multidimensional blocks whose internal coordinates are not uniquely determined. When such blocks are used to represent and intervene on higher-level concepts, it is unclear whether th...
Training-free diffusion-based compression methods repurpose the generative priors of a diffusion or a flow model to compress data without retraining them as task-specific codecs. Recent codebook-based approaches such as Turbo-DDCM achieve strong perceptual quality at ultra-low bitrates, but their reliance on dense Gaus...
Or Ronai, Vladimir Kulikov, T. Michaeli· 0 citations
{\em Specification-Guided Reinforcement Learning (RL)} has largely concerned itself with \emph{learning} an optimal policy against a formal specification in an unknown {\em Markov Decision Process (MDP)}. We take up the dual question. Given a policy already in hand, can one \emph{verify} that it is correct against a fo...
S. Akshay, Suguman Bansal, Gaurav Mahla· 0 citations
Graph Neural Networks (GNNs) commonly assume that node features are fully observed, despite missing features being pervasive in real-world applications. By leveraging the dependence between node features and graph structure, existing graph-specific methods handle missingness by propagating observed feature values along...
Enrico Chial\`a, Francesco Ferrini, Antonio Longa et al.· 0 citations
Large language models (LLMs) are increasingly used as judges to predict which of two candidate answers a human evaluator would prefer, providing a scalable alternative when human evaluation is too costly. Unfortunately, LLM judgments can exhibit systematic biases. Prior work has documented position, verbosity or model...
Cesare Barbera, Luca Stradiotti, Lorenzo Cascioli et al.· 0 citations
Post-training is crucial for improving text-to-image models, yet existing pipelines typically rely on manually curated training data and fixed training distributions. We introduce APT (Agentic Post-Training), a fully automated framework that adaptively post-trains a base model according to its evolving capability profi...
Si-Rui Zhang, Fang-Yun Wei, Dong Chen et al.· 0 citations
Reliable uncertainty evaluation is essential for assessing the trustworthiness of machine learning (ML) predictions, particularly in safety-critical applications. This paper explores conceptual analogies between uncertainty evaluation in ML and measurement science, drawing on the evolution of metrology from error-based...
Transformers can implement reinforcement-learning algorithms in context, yet how off-policy control emerges within Transformers remains poorly understood. We study Q-learning, a foundational model-free control algorithm that combines off-policy bootstrapping with greedy policy improvement. We prove that a two-layer att...
