Weekly Top 10 · W47

2026-11-16 ~ 11-22 · Infra

Generative · Personalization · AI Search · Infra · Embodied AI

  1. 1. Advancing Quantization for R&R Models with Microscaling Formats · 📖 Deep dive

    Ads ranking models are extremely sensitive to quantization error. A launch bar often requires that normalized entropy stay within 0.01% of the baseline, much tighter than the accuracy loss large language models often…

    Appeared in: 11-17

  2. 2. Authenticated Approximate Nearest Neighbor Search in Outsourced Vector Databases · 📖 Deep dive

    Once a vector database is outsourced to a third-party cloud, the provider may skip neighbors to save compute or even tamper with results, so the client needs a way to verify that a nearest-neighbor search is genuine…

    Appeared in: 11-16

  3. 3. Benchmarking tiered KV cache offloading on H100 GPUs in OpenShift AI 3.5

    Long-context agent requests resend the same prefix again and again, so the KV cache could be reused, but once GPU memory fills up it is thrown away and recomputed next time…

    Appeared in: 11-16

  4. 4. SoScale: Second-Order Scale Search for FP4 Quantization · 📖 Deep dive

    In FP4 quantization each number has only 16 possible values, and each block of data gets one scale that decides where those grid points land. The default scales by the block's maximum absolute value…

    Appeared in: 11-16

  5. 5. VecDBLens: An Evaluatology-Based Framework for Vector Database Diagnosis · 📖 Deep dive

    When a vector database recalls poorly, the first instinct is to tune the index, but the ceiling on what an index can find is set by the embedding. On VisualNews with SigLIP2 and exact brute-force search…

    Appeared in: 11-16

  6. 6. TIDE: Reclaiming Lost Performance In On-Policy Reinforcement Learning · 📖 Deep dive

    In synchronous, colocated reinforcement learning, inference engines often sit idle waiting for a few very long generations, or decode inefficiently at very small batch sizes, and GPUs go underused…

    Appeared in: 11-17

  7. 7. Compiling Specialized GPU-Initiated Networking · 📖 Deep dive

    Machine learning systems often specialize computation to the workload and the hardware, but communication still lives in generic libraries, whose device code and transport choices a high-level compiler cannot see…

    Appeared in: 11-17

  8. 8. Zero-SM NCCL: A Systematic Guide to Designing and Deploying SM-Free GPU Communication · 📖 Deep dive

    GPU communication libraries often use compute units to move data and synchronize, and the cost of stealing compute grows as interconnect bandwidth rises. The authors instead complete collectives with copy engines and…

    Appeared in: 11-17

  9. 9. GC-MoE on SM120: Overlapping Expert-Parallel Token Dispatch with Warp-Level FP8xFP4 Block-Scaled MMA on Consumer Blackwell

    Expert-parallel MoE decoding is latency bound: every layer must dispatch tokens to the GPUs, run grouped matrix multiplies over the experts, and combine the results, three steps in series…

    Appeared in: 11-16

  10. 10. Snoopie: A Multi-GPU Communication Profiler and Visualizer

    Multi-GPU programs often fail to scale because of data movement, yet existing tools cannot show clearly who moves what. NVIDIA's Nsight Systems only hints that there is traffic on NVLink…

    Appeared in: 11-16

‹ Previous week