Picture for Philip H. S. Torr

Philip H. S. Torr

University of Oxford

Prompting a Pretrained Transformer Can Be a Universal Approximator

Add code
Feb 22, 2024
Viaarxiv icon

Self-consistent Validation for Machine Learning Electronic Structure

Add code
Feb 15, 2024
Viaarxiv icon

Revealing Decurve Flows for Generalized Graph Propagation

Add code
Feb 13, 2024
Viaarxiv icon

RanDumb: A Simple Approach that Questions the Efficacy of Continual Representation Learning

Add code
Feb 13, 2024
Figure 1 for RanDumb: A Simple Approach that Questions the Efficacy of Continual Representation Learning
Figure 2 for RanDumb: A Simple Approach that Questions the Efficacy of Continual Representation Learning
Figure 3 for RanDumb: A Simple Approach that Questions the Efficacy of Continual Representation Learning
Figure 4 for RanDumb: A Simple Approach that Questions the Efficacy of Continual Representation Learning
Viaarxiv icon

Secret Collusion Among Generative AI Agents

Add code
Feb 12, 2024
Figure 1 for Secret Collusion Among Generative AI Agents
Figure 2 for Secret Collusion Among Generative AI Agents
Figure 3 for Secret Collusion Among Generative AI Agents
Figure 4 for Secret Collusion Among Generative AI Agents
Viaarxiv icon

From Categories to Classifier: Name-Only Continual Learning by Exploring the Web

Add code
Nov 19, 2023
Viaarxiv icon

When Do Prompting and Prefix-Tuning Work? A Theory of Capabilities and Limitations

Add code
Oct 30, 2023
Viaarxiv icon

Revisiting Evaluation Metrics for Semantic Segmentation: Optimization and Evaluation of Fine-grained Intersection over Union

Add code
Oct 30, 2023
Viaarxiv icon

Segment, Select, Correct: A Framework for Weakly-Supervised Referring Segmentation

Add code
Oct 23, 2023
Viaarxiv icon

Fine-tuning can cripple your foundation model; preserving features may be the solution

Add code
Aug 25, 2023
Viaarxiv icon