Picture for Tianyi Zhou

Tianyi Zhou

The Crystal Ball Hypothesis in diffusion models: Anticipating object positions from initial noise

Add code
Jun 04, 2024
Figure 1 for The Crystal Ball Hypothesis in diffusion models: Anticipating object positions from initial noise
Figure 2 for The Crystal Ball Hypothesis in diffusion models: Anticipating object positions from initial noise
Figure 3 for The Crystal Ball Hypothesis in diffusion models: Anticipating object positions from initial noise
Figure 4 for The Crystal Ball Hypothesis in diffusion models: Anticipating object positions from initial noise
Viaarxiv icon

Enhancing Visual-Language Modality Alignment in Large Vision Language Models via Self-Improvement

Add code
May 29, 2024
Figure 1 for Enhancing Visual-Language Modality Alignment in Large Vision Language Models via Self-Improvement
Figure 2 for Enhancing Visual-Language Modality Alignment in Large Vision Language Models via Self-Improvement
Figure 3 for Enhancing Visual-Language Modality Alignment in Large Vision Language Models via Self-Improvement
Figure 4 for Enhancing Visual-Language Modality Alignment in Large Vision Language Models via Self-Improvement
Viaarxiv icon

Multi-Level Additive Modeling for Structured Non-IID Federated Learning

Add code
May 26, 2024
Viaarxiv icon

Mosaic IT: Enhancing Instruction Tuning with Data Mosaics

Add code
May 22, 2024
Viaarxiv icon

Adaptive Regularization of Representation Rank as an Implicit Constraint of Bellman Equation

Add code
Apr 19, 2024
Viaarxiv icon

Simplicity Bias of Transformers to Learn Low Sensitivity Functions

Add code
Mar 11, 2024
Viaarxiv icon

Many-Objective Multi-Solution Transport

Add code
Mar 06, 2024
Viaarxiv icon

DrAttack: Prompt Decomposition and Reconstruction Makes Powerful LLM Jailbreakers

Add code
Mar 01, 2024
Figure 1 for DrAttack: Prompt Decomposition and Reconstruction Makes Powerful LLM Jailbreakers
Figure 2 for DrAttack: Prompt Decomposition and Reconstruction Makes Powerful LLM Jailbreakers
Figure 3 for DrAttack: Prompt Decomposition and Reconstruction Makes Powerful LLM Jailbreakers
Figure 4 for DrAttack: Prompt Decomposition and Reconstruction Makes Powerful LLM Jailbreakers
Viaarxiv icon

Meta-Task Prompting Elicits Embedding from Large Language Models

Add code
Feb 28, 2024
Viaarxiv icon

Corpus-Steered Query Expansion with Large Language Models

Add code
Feb 28, 2024
Viaarxiv icon