↓Skip to main content

🏢 University College London

Structured Learning of Compositional Sequential Interventions
·2166 words·11 mins· loading · loading
AI Generated AI Theory Causality 🏢 University College London
Predicting outcomes of combined sequential interventions is challenging, especially in sparse data. This paper introduces CSI-VAE, a novel compositional model that provides reliable predictions for u…
Sample-efficient Bayesian Optimisation Using Known Invariances
·1812 words·9 mins· loading · loading
AI Theory Optimization 🏢 University College London
Boost Bayesian Optimization’s efficiency by leveraging known invariances in objective functions for faster, more effective solutions.
ReplaceAnything3D: Text-Guided Object Replacement in 3D Scenes with Compositional Scene Representations
·4073 words·20 mins· loading · loading
AI Generated Multimodal Learning Vision-Language Models 🏢 University College London
ReplaceAnything3D (RAM3D) revolutionizes 3D scene editing with a text-guided, multi-view consistent approach for seamlessly replacing or adding 3D objects in complex scenes.
Reparameterized Multi-Resolution Convolutions for Long Sequence Modelling
·3179 words·15 mins· loading · loading
Machine Learning Deep Learning 🏢 University College London
MRConv: Reparameterized multi-resolution convolutions efficiently model long sequences, improving performance across various data modalities.
REDUCR: Robust Data Downsampling using Class Priority Reweighting
·2544 words·12 mins· loading · loading
Machine Learning Deep Learning 🏢 University College London
REDUCR, a novel data downsampling method, significantly improves worst-class test accuracy in imbalanced datasets by using class priority reweighting, surpassing state-of-the-art methods by ~15%.
Practical Shuffle Coding
·2298 words·11 mins· loading · loading
Machine Learning Deep Learning 🏢 University College London
Revolutionizing unordered data compression, this paper introduces autoregressive shuffle coding, achieving state-of-the-art speeds and compression rates on massive datasets.
Nearly Tight Black-Box Auditing of Differentially Private Machine Learning
·1819 words·9 mins· loading · loading
AI Theory Privacy 🏢 University College London
This paper presents a new auditing method for DP-SGD that provides substantially tighter black-box privacy analyses than previous methods, yielding significantly closer empirical estimates to theoreti…
Instruction Tuning With Loss Over Instructions
·4022 words·19 mins· loading · loading
AI Generated Natural Language Processing Large Language Models 🏢 University College London
Boost LLM performance with INSTRUCTION MODELLING: a simple yet effective instruction tuning method that improves model outputs by over 100% in some cases by applying loss to both instructions and outp…
DEFT: Efficient Fine-tuning of Diffusion Models by Learning the Generalised $h$-transform
·3837 words·19 mins· loading · loading
AI Generated Machine Learning Deep Learning 🏢 University College London
DEFT: A novel method efficiently fine-tunes diffusion models for conditional generation via a generalized h-transform, achieving state-of-the-art performance with significant speed improvements.
Controlling Multiple Errors Simultaneously with a PAC-Bayes Bound
·547 words·3 mins· loading · loading
AI Generated AI Theory Generalization 🏢 University College London
New PAC-Bayes bound controls multiple error types simultaneously, providing richer generalization guarantees.
Adversarially Robust Decision Transformer
·2778 words·14 mins· loading · loading
AI Theory Robustness 🏢 University College London
Adversarially Robust Decision Transformer (ARDT) enhances offline RL robustness against powerful adversaries by conditioning policies on minimax returns, achieving superior worst-case performance.