🏗️ Building on HF

Haotian Ma PRO

Hunter26666666

https://atheavenma.github.io/

AI & ML interests

I develop methods to understand how learned representations shape model behavior across interventions, combinations, and changing contexts.

Recent Activity

updated a collection 4 days ago

Interpretable AI for Scientific Discovery

upvoted a paper 4 days ago

AxBench: Steering LLMs? Even Simple Baselines Outperform Sparse Autoencoders

upvoted a collection 4 days ago

Interpretable AI for Scientific Discovery

View all activity

Organizations

upvoted a paper 4 days ago

AxBench: Steering LLMs? Even Simple Baselines Outperform Sparse Autoencoders

Paper • 2501.17148 • Published Jan 28, 2025 • 2

upvoted a collection 4 days ago

Interpretable AI for Scientific Discovery

Collection

Develop interpretable methods that connect model decisions to scientifically meaningful spatial, spectral, and molecular structure. • 3 items • Updated 4 days ago • 1

upvoted a paper 6 days ago

Representation Redistribution in Sparse Autoencoders

Paper • 2610.21240 • Published 19 days ago • 10

upvoted a collection 12 days ago

Faithful Representations and Mechanistic Interpretability

Collection

Understand feature representations and how those representations support faithful explanations and interventions. • 3 items • Updated 4 days ago • 1

upvoted a paper 12 days ago

Spectral-Spatial Interpretation

Paper • 2610.21114 • Published 19 days ago • 62

upvoted 2 collections 17 days ago

SAE Benchmarks, Datasets & Models

Collection

Benchmarks, datasets, and pretrained models for reproducible sparse autoencoder research, feature interventions, and mechanistic interpretability. • 11 items • Updated 14 days ago • 1

Behavioral Compositionality of SAE Representations

Collection

Research, benchmarks, and tools for measuring how sparse autoencoder features interact to shape model behavior. • 2 items • Updated 14 days ago • 1

upvoted a paper 18 days ago

Behavioral Compositionality of Sparse Autoencoder Representations

Paper • 2610.21261 • Published 19 days ago • 26

upvoted an article 18 days ago

Article

Scaling Indic Pretraining: 2.8 Trillion Tokens and an Open Data Generation Pipeline

nvidia

•

18 days ago

• 50

Haotian Ma PRO

AI & ML interests

Recent Activity

Organizations

Hunter26666666's activity

Scaling Indic Pretraining: 2.8 Trillion Tokens and an Open Data Generation Pipeline