Blogs on Hao AI Lab @ UCSD haoailab.com

Attn-QAT: Making 4-Bit Attention Actually Work

haoailab.com·6w

placeholder: Interactive Vibe Directing in FastVideo

haoailab.com·9w

FastLTX-2.3: The Fastest and Cheapest 1080p TI2AV Pipeline Ever

haoailab.com·10w

From Physical Commonsense to Scientific Reasoning: Why World Modeling in Video Matters

hao-ai-lab.github.io·14w

CAD: Disaggregating Core Attention for Efficient Long-context Language Model Training

hao-ai-lab.github.io·126w·Hacker News

Fast and Accurate Causal Parallel Decoding using Jacobi Forcing

hao-ai-lab.github.io·22w

AUP: when Accuracy Meets Parallelism in Diffusion Language Models

hao-ai-lab.github.io·23w

CausalWan-MoE Preview: Applying Self-Forcing Distillation To Wan2.2

hao-ai-lab.github.io·26w·Hacker News

Disaggregated Inference: The Past, Present and Future

hao-ai-lab.github.io·30w·Hacker News

Scaling Speculative Decoding with Lookahead Reasoning

hao-ai-lab.github.io·34w

Can RL-based LLM post-training on games generalize to other tasks? (GRL)

hao-ai-lab.github.io·38w

A Practical Guideline to Using Lmgame-Bench

hao-ai-lab.github.io·39w

Introducing FastWAN: 15 Times Faster Video Generation With Sparse Distillation

hao-ai-lab.github.io·42w·Hacker News

From Pokémon Red to Standardized Game-as-an-Eval

hao-ai-lab.github.io·48w

FastVideo V1: A Unified Framework for Accelerated Video Generation

hao-ai-lab.github.io·56w

ReFoRCE: A Text-to-SQL Agent with Self-Refinement, Format Restriction, and Column Exploration

hao-ai-lab.github.io·58w

Fast Video Generation with Sliding Tile Attention

hao-ai-lab.github.io·65w·Hacker News

Reasoning Without Hesitating: Efficient Cot Through Certainty Probing

hao-ai-lab.github.io·65w·Hacker News

GameArena: Evaluating LLM Reasoning through Live Computer Games

hao-ai-lab.github.io·66w

Efficient LLM Scheduling by Learning to Rank

hao-ai-lab.github.io·70w·Hacker News

Log in to enable infinite scrolling