Picture for Alessio Devoto

Alessio Devoto

Q-Filters: Leveraging QK Geometry for Efficient KV Cache Compression

Add code
Mar 04, 2025
Viaarxiv icon

Mixture-of-Experts Graph Transformers for Interpretable Particle Collision Detection

Add code
Jan 08, 2025
Figure 1 for Mixture-of-Experts Graph Transformers for Interpretable Particle Collision Detection
Figure 2 for Mixture-of-Experts Graph Transformers for Interpretable Particle Collision Detection
Figure 3 for Mixture-of-Experts Graph Transformers for Interpretable Particle Collision Detection
Figure 4 for Mixture-of-Experts Graph Transformers for Interpretable Particle Collision Detection
Viaarxiv icon

Goal-oriented Communications based on Recursive Early Exit Neural Networks

Add code
Dec 27, 2024
Viaarxiv icon

Analysing the Residual Stream of Language Models Under Knowledge Conflicts

Add code
Oct 21, 2024
Figure 1 for Analysing the Residual Stream of Language Models Under Knowledge Conflicts
Figure 2 for Analysing the Residual Stream of Language Models Under Knowledge Conflicts
Figure 3 for Analysing the Residual Stream of Language Models Under Knowledge Conflicts
Figure 4 for Analysing the Residual Stream of Language Models Under Knowledge Conflicts
Viaarxiv icon

Steering Knowledge Selection Behaviours in LLMs via SAE-Based Representation Engineering

Add code
Oct 21, 2024
Figure 1 for Steering Knowledge Selection Behaviours in LLMs via SAE-Based Representation Engineering
Figure 2 for Steering Knowledge Selection Behaviours in LLMs via SAE-Based Representation Engineering
Figure 3 for Steering Knowledge Selection Behaviours in LLMs via SAE-Based Representation Engineering
Figure 4 for Steering Knowledge Selection Behaviours in LLMs via SAE-Based Representation Engineering
Viaarxiv icon

Adaptive Layer Selection for Efficient Vision Transformer Fine-Tuning

Add code
Aug 16, 2024
Figure 1 for Adaptive Layer Selection for Efficient Vision Transformer Fine-Tuning
Figure 2 for Adaptive Layer Selection for Efficient Vision Transformer Fine-Tuning
Figure 3 for Adaptive Layer Selection for Efficient Vision Transformer Fine-Tuning
Figure 4 for Adaptive Layer Selection for Efficient Vision Transformer Fine-Tuning
Viaarxiv icon

A Simple and Effective $L_2$ Norm-Based Strategy for KV Cache Compression

Add code
Jun 17, 2024
Figure 1 for A Simple and Effective $L_2$ Norm-Based Strategy for KV Cache Compression
Figure 2 for A Simple and Effective $L_2$ Norm-Based Strategy for KV Cache Compression
Figure 3 for A Simple and Effective $L_2$ Norm-Based Strategy for KV Cache Compression
Figure 4 for A Simple and Effective $L_2$ Norm-Based Strategy for KV Cache Compression
Viaarxiv icon

Are We Done with MMLU?

Add code
Jun 07, 2024
Figure 1 for Are We Done with MMLU?
Figure 2 for Are We Done with MMLU?
Figure 3 for Are We Done with MMLU?
Figure 4 for Are We Done with MMLU?
Viaarxiv icon

Adaptive Semantic Token Selection for AI-native Goal-oriented Communications

Add code
Apr 25, 2024
Viaarxiv icon

Conditional computation in neural networks: principles and research trends

Add code
Mar 12, 2024
Figure 1 for Conditional computation in neural networks: principles and research trends
Figure 2 for Conditional computation in neural networks: principles and research trends
Figure 3 for Conditional computation in neural networks: principles and research trends
Figure 4 for Conditional computation in neural networks: principles and research trends
Viaarxiv icon