What we are reading: paper pills
Paper pills are summaries of publications, talks or new software, with context and a short description of the methods, main ideas and key results. They are written for practitioners or researchers who want a peek into recent developments but are short on time. We publish monthly digests as well as highlights of major conferences.

Pills

Deep Reinforcement Learning at the Edge of the Statistical Precipice

A sound statistical analysis of the performance of RL algorithms is performed, with outcomes that force a reconsideration of the ranking of several recently published methods. The important call for action for more careful statistical reasoning in the reporting of results that was formulated in this paper received an Outstanding Paper Award at NeurIPS21.

A picture of Michael
Efficient RL

Beta Shapley: a Unified and Noise-reduced Data Valuation Framework for Machine Learning

In data valuation, relaxing the Shapley axioms leads to weighting schemes for marginal utilities that improve the quality of values

A picture of Miguel
Data Valuation

Planning with Diffusion for Flexible Behavior Synthesis

Diffusion models are finding their way into more machine learning domains. This paper introduces a framework for applying them to model-based reinforcement learning.

A photo of Fabio
Efficient RL
Diffusion Models

Prompt-to-Prompt Image Editing with Cross Attention Control

Generating images from text is complicated by the fact that small changes in prompt gives very different results. This new study shows how to do minimal changes to generated images through manipulation of the neural network attention maps.

A photo of Fabio
Diffusion Models

Optimization Based Modelling

Optimization layers allow incorporating downstream optimization problems into neural network training. The performance gap between the first-learn-then-optimise and the end-to-end approaches in the perfect model setting is shown to be unbounded for non-linear cost functions.

A picture of Faried

DictBERT: Dictionary Description Knowledge Enhanced Language Model Pre-training via Contrastive Learning

External knowledge can be injected into pre-trained language models (here specifically BERT) by training an additional language model on various dictionary-related tasks. This strategy may result in better representations and should be especially useful for sentences with abundant jargon.

A picture of Michael
Large Language Models

Local calibration: metrics and recalibration

A calibration method that takes sample similarity into account, automatically providing group calibration even when groups are unknown.

A picture of Miguel
Classifier Calibration

Towards Generative Modelling of Multivariate Extremes and Tail Dependence

Modeling heavy-tailed marginal distributions and asymmetric tail dependence with normalizing flows and copula theory.

A picture of Faried

Neuro Symbolic Verification of Deep Neural Networks

The neuro-symbolic approach to verification of neural networks uses reference networks to represent high level concepts. This allows the application of formal verification techniques to deep learning.

A picture of Faried
Formal Verification Under Uncertainty

Probabilistic Inference With Algebra and Logic Constraints

Hybrid probabilistic inference allows performing probabilistic inference with algebra and logic constraints. Recent advances with emphasis on computational tractability are reviewed and compared against each other.

A picture of Faried
Formal Verification Under Uncertainty