Picture for Jiantao Jiao

Jiantao Jiao

How Do LLMs Perform Two-Hop Reasoning in Context?

Add code
Feb 19, 2025
Viaarxiv icon

Token Assorted: Mixing Latent and Text Tokens for Improved Language Model Reasoning

Add code
Feb 05, 2025
Figure 1 for Token Assorted: Mixing Latent and Text Tokens for Improved Language Model Reasoning
Figure 2 for Token Assorted: Mixing Latent and Text Tokens for Improved Language Model Reasoning
Figure 3 for Token Assorted: Mixing Latent and Text Tokens for Improved Language Model Reasoning
Figure 4 for Token Assorted: Mixing Latent and Text Tokens for Improved Language Model Reasoning
Viaarxiv icon

How to Evaluate Reward Models for RLHF

Add code
Oct 18, 2024
Figure 1 for How to Evaluate Reward Models for RLHF
Figure 2 for How to Evaluate Reward Models for RLHF
Figure 3 for How to Evaluate Reward Models for RLHF
Figure 4 for How to Evaluate Reward Models for RLHF
Viaarxiv icon

Active-Dormant Attention Heads: Mechanistically Demystifying Extreme-Token Phenomena in LLMs

Add code
Oct 17, 2024
Figure 1 for Active-Dormant Attention Heads: Mechanistically Demystifying Extreme-Token Phenomena in LLMs
Figure 2 for Active-Dormant Attention Heads: Mechanistically Demystifying Extreme-Token Phenomena in LLMs
Figure 3 for Active-Dormant Attention Heads: Mechanistically Demystifying Extreme-Token Phenomena in LLMs
Figure 4 for Active-Dormant Attention Heads: Mechanistically Demystifying Extreme-Token Phenomena in LLMs
Viaarxiv icon

Thinking LLMs: General Instruction Following with Thought Generation

Add code
Oct 14, 2024
Figure 1 for Thinking LLMs: General Instruction Following with Thought Generation
Figure 2 for Thinking LLMs: General Instruction Following with Thought Generation
Figure 3 for Thinking LLMs: General Instruction Following with Thought Generation
Figure 4 for Thinking LLMs: General Instruction Following with Thought Generation
Viaarxiv icon

EmbedLLM: Learning Compact Representations of Large Language Models

Add code
Oct 03, 2024
Viaarxiv icon

Meta-Rewarding Language Models: Self-Improving Alignment with LLM-as-a-Meta-Judge

Add code
Jul 28, 2024
Figure 1 for Meta-Rewarding Language Models: Self-Improving Alignment with LLM-as-a-Meta-Judge
Figure 2 for Meta-Rewarding Language Models: Self-Improving Alignment with LLM-as-a-Meta-Judge
Figure 3 for Meta-Rewarding Language Models: Self-Improving Alignment with LLM-as-a-Meta-Judge
Figure 4 for Meta-Rewarding Language Models: Self-Improving Alignment with LLM-as-a-Meta-Judge
Viaarxiv icon

Universal evaluation and design of imaging systems using information estimation

Add code
May 31, 2024
Figure 1 for Universal evaluation and design of imaging systems using information estimation
Figure 2 for Universal evaluation and design of imaging systems using information estimation
Figure 3 for Universal evaluation and design of imaging systems using information estimation
Figure 4 for Universal evaluation and design of imaging systems using information estimation
Viaarxiv icon

Toxicity Detection for Free

Add code
May 29, 2024
Viaarxiv icon

Toward a Theory of Tokenization in LLMs

Add code
Apr 12, 2024
Figure 1 for Toward a Theory of Tokenization in LLMs
Figure 2 for Toward a Theory of Tokenization in LLMs
Figure 3 for Toward a Theory of Tokenization in LLMs
Figure 4 for Toward a Theory of Tokenization in LLMs
Viaarxiv icon