Related papers: Decomposing the Generalization Gap in PROTAC Activ…
When evaluating medical tests or biomarkers for disease classification, the area under the receiver-operating characteristic (ROC) curve is a widely used performance metric that does not require us to commit to a specific decision…
We explore colour versus shape goal misgeneralization originally demonstrated by Di Langosco et al. (2022) in the Procgen Maze environment, where, given an ambiguous choice, the agents seem to prefer generalization based on colour rather…
Human body consists of lot of cells, each cell consist of DeOxaRibo Nucleic Acid (DNA). Identifying the genes from the DNA sequences is a very difficult task. But identifying the coding regions is more complex task compared to the former.…
The phenomena of stochasticity in biochemical processes have been intriguing life scientists for the past few decades. We now know that living cells take advantage of stochasticity in some cases and counteract stochastic effects in others.…
Large Language Models (LLMs) have shown impressive performance as general purpose agents, but their abilities remain highly dependent on prompts which are hand written with onerous trial-and-error effort. We propose a simple and…
Background: Despite high in-silico performance (AUC >0.80), 85% of AI cancer biomarkers fail clinical translation, exposing a critical algorithm-to-outcome gap. Methods: We introduce the Algorithm-to-Outcome Concordance (AOC) framework,…
We give a general recipe for derandomising PAC-Bayesian bounds using margins, with the critical ingredient being that our randomised predictions concentrate around some value. The tools we develop straightforwardly lead to margin bounds for…
Cross-dose denoising for low-dose positron emission tomography (LDPET) has been proposed to address the limited generalization of models trained at a single noise level. However, neural networks trained on a specific dose level often fail…
Recently, Vision-Language foundation models like CLIP and ALIGN, which are pre-trained on large-scale data have shown remarkable zero-shot generalization to diverse datasets with different classes and even domains. In this work, we take a…
Predictive coding networks are neural models that perform inference through an iterative energy minimization process, whose operations are local in space and time. While effective in shallow architectures, they suffer significant…
Interleaving is an online evaluation approach for information retrieval systems that compares the effectiveness of ranking functions in interpreting the users' implicit feedback. Previous work such as Hofmann et al (2011) has evaluated the…
Embodied AI agents exploit reward signal flaws through reward hacking, achieving high proxy scores while failing true objectives. We introduce Mechanistically Interpretable Task Decomposition (MITD), a hierarchical transformer architecture…
We study the problem of learning-to-learn: inferring a learning algorithm that works well on tasks sampled from an unknown distribution. As class of algorithms we consider Stochastic Gradient Descent on the true risk regularized by the…
A multi-reader multi-case (MRMC) analysis is applied to account for both reader and case variability when evaluating the clinical performance of a medical imaging device or reader performance under different reading modalities. For a…
Pretrained vision-language models (VLMs) such as CLIP have shown impressive generalization capability in downstream vision tasks with appropriate text prompts. Instead of designing prompts manually, Context Optimization (CoOp) has been…
We introduce a proxy-analyzer framework for detecting hallucinations in large language models. Instead of looking inside the generating model, our system reads already-generated text through a small locally hosted open-weight model and…
Extreme low-data fine-grained classification is common in expert domains where labeling is expensive, yet practitioners still need principled guidance for selecting pretrained encoders. We study emerald inclusion grading with a custom…
Activity and property prediction models are the central workhorses in drug discovery and materials sciences, but currently they have to be trained or fine-tuned for new tasks. Without training or fine-tuning, scientific language models…
Advances in hardware and language model architecture have spurred a revolution in natural language generation. However, autoregressive models compute probability distributions over next-token choices, and sampling from these distributions,…
This paper proposes a new method for estimating conditional average treatment effects (CATE) in randomized experiments. We adopt inverse probability weighting (IPW) for identification; however, IPW-transformed outcomes are known to be…