Stories by fzliu

voyage-multimodal-3.5: a new multimodal retrieval frontier with video support

The Voyage 4 model family: shared embedding space with MoE architecture

Token-Count-Based Batching: Faster, Cheaper Embedding Inference for Queries

The Case Against LLMs as Rerankers

Ultra-Wide Band: A Transformational Technology for the Internet of Things

Token-Count-Based Batching: Faster, Cheaper Embedding Inference for Queries

Mapping China's Surnames

Token-Count-Based Batching: Faster, Cheaper Embedding Inference for Queries

How geometry is fundamental for chess

Nimzo – Grischuk Gambit

Threatening AI models has no meaningful effect on performance

Do Attacking Players Play More Forcing Moves?

The Case Against LLMs as Rerankers

Nearest Neighbor Speculative Decoding for LLM Generation and Attribution

Restructuring Vector Quantization with the Rotation Trick

There Are No Good Anti-Scandinavians

The Geographic Distribution of China's Last Names, in Maps (2013)

Think Faster, Talk Smarter with Matt Abrahams [video]

The Case Against LLMs as Rerankers

The next RISC-V processor frontier: AI

1 2 3 4 5 6 7 8 9 10