Picture for Brando Miranda

Brando Miranda

When Do Universal Image Jailbreaks Transfer Between Vision-Language Models?

Add code
Jul 21, 2024
Viaarxiv icon

Why Has Predicting Downstream Capabilities of Frontier AI Models with Scale Remained Elusive?

Add code
Jun 06, 2024
Figure 1 for Why Has Predicting Downstream Capabilities of Frontier AI Models with Scale Remained Elusive?
Figure 2 for Why Has Predicting Downstream Capabilities of Frontier AI Models with Scale Remained Elusive?
Figure 3 for Why Has Predicting Downstream Capabilities of Frontier AI Models with Scale Remained Elusive?
Figure 4 for Why Has Predicting Downstream Capabilities of Frontier AI Models with Scale Remained Elusive?
Viaarxiv icon

An Evaluation Benchmark for Autoformalization in Lean4

Add code
Jun 01, 2024
Viaarxiv icon

Is Pre-training Truly Better Than Meta-Learning?

Add code
Jun 24, 2023
Viaarxiv icon

Beyond Scale: the Diversity Coefficient as a Data Quality Metric Demonstrates LLMs are Pre-trained on Formally Diverse Data

Add code
Jun 24, 2023
Viaarxiv icon

Are Emergent Abilities of Large Language Models a Mirage?

Add code
Apr 28, 2023
Viaarxiv icon

The Curse of Low Task Diversity: On the Failure of Transfer Learning to Outperform MAML and Their Empirical Equivalence

Add code
Aug 02, 2022
Figure 1 for The Curse of Low Task Diversity: On the Failure of Transfer Learning to Outperform MAML and Their Empirical Equivalence
Figure 2 for The Curse of Low Task Diversity: On the Failure of Transfer Learning to Outperform MAML and Their Empirical Equivalence
Figure 3 for The Curse of Low Task Diversity: On the Failure of Transfer Learning to Outperform MAML and Their Empirical Equivalence
Figure 4 for The Curse of Low Task Diversity: On the Failure of Transfer Learning to Outperform MAML and Their Empirical Equivalence
Viaarxiv icon

Does MAML Only Work via Feature Re-use? A Data Centric Perspective

Add code
Dec 24, 2021
Figure 1 for Does MAML Only Work via Feature Re-use? A Data Centric Perspective
Figure 2 for Does MAML Only Work via Feature Re-use? A Data Centric Perspective
Figure 3 for Does MAML Only Work via Feature Re-use? A Data Centric Perspective
Figure 4 for Does MAML Only Work via Feature Re-use? A Data Centric Perspective
Viaarxiv icon

The Curse of Zero Task Diversity: On the Failure of Transfer Learning to Outperform MAML and their Empirical Equivalence

Add code
Dec 24, 2021
Figure 1 for The Curse of Zero Task Diversity: On the Failure of Transfer Learning to Outperform MAML and their Empirical Equivalence
Figure 2 for The Curse of Zero Task Diversity: On the Failure of Transfer Learning to Outperform MAML and their Empirical Equivalence
Figure 3 for The Curse of Zero Task Diversity: On the Failure of Transfer Learning to Outperform MAML and their Empirical Equivalence
Viaarxiv icon

Theory III: Dynamics and Generalization in Deep Networks - a simple solution

Add code
Apr 11, 2019
Figure 1 for Theory III: Dynamics and Generalization in Deep Networks - a simple solution
Figure 2 for Theory III: Dynamics and Generalization in Deep Networks - a simple solution
Figure 3 for Theory III: Dynamics and Generalization in Deep Networks - a simple solution
Figure 4 for Theory III: Dynamics and Generalization in Deep Networks - a simple solution
Viaarxiv icon