Criticizing dot product retrieval, advocating for inference scaling
lateinteraction · x · 2026-08-26
The author argues that in ML, retrieval tasks still rely on inert scoring functions (like the simple dot product), forcing the use of exponentially larger embeddings for basic improvements. The author advocates for "inference scaling for retrieval," also known as Late Interaction from 2019/2020. Following this, the author hinted at taking matters into their own hands to implement these ideas due to the inertia in the field.
More from Research
- OpenAI staff claims Transformer is now the bottleneck to AI progress — KyeGomezB · 2026-08-26
- Alibaba's Qwen World Model: Mental Simulation for AI Agents — 大模型之路 · 2026-08-26
- Latency reduction in MoE models comes at high cost — firstadopter · 2026-08-26
- Artificial Analysis Introduces Reward Hacking Corrections to Coding Agent Index — ArtificialAnlys · 2026-08-26
- Qwen3.8-27B runs 100k-token physics simulation on a 16GB GPU — 1000_bucks_a_month · 2026-08-26
- Study: Can AI Antibody Design Beat CDR Consensus? — anshulkundaje · 2026-08-26