Skip to content
TodayNov 17, 2026Tue2items
Nov 12, 2026Thu
  1. Meta Engineering · ML Applications42

    Meta details Intelligent Delivery system for allocating ad-ranking compute

    Meta Engineering described Intelligent Delivery, an ML-driven system that allocates compute across its ads serving stack toward a system-wide optimum. It has delivered about 10% incremental ranking performance on top of model and system advancements from 2017 through the first half of 2026. In that half, AI agents contributed to roughly 20% of its optimizations.

Nov 11, 2026Wed
  1. Meta Engineering · ML Applications58

    Meta details TLX, a Triton extension for explicit GPU warp-group control

    Meta Engineering introduced Triton Low-Level Extensions (TLX), which adds explicit warp-group and cluster-level orchestration to Triton while keeping its programming model. TLX targets NVIDIA Hopper and Blackwell and AMD MI300/MI350x. Meta reports 20-25% end-to-end MFU on GEM training across thousands of Blackwell GPUs, up to two-thirds lower compute-intensive latency in RecSys inference, and production use in LLM pretraining.

Nov 10, 2026Tue
  1. Amazon Science60

    Amazon releases Mitra-v2 tabular foundation model, open-sourced on Hugging Face

    Amazon Science introduced Mitra-v2, a second-generation tabular foundation model for classification and regression, pretrained entirely on synthetic tables with 10 times as many rows as Mitra-v1. It leads the single-model field on TabArena's 51 real-world datasets and ranks first on ScoringBench for regression. Fine-tuned models and inference code are released under Apache-2.0 on Hugging Face.

    Why it matters: Open, synthetic-only pretraining that tops public tabular benchmarks could reduce per-table feature engineering and tuning, a common cost in enterprise ML pipelines.

Oct 23, 2026Fri
  1. Amazon Science58

    Amazon releases TorchNeuron, an open-source PyTorch backend for AWS Trainium

    Amazon announced TorchNeuron, an open-source backend that translates PyTorch code into operations runnable on AWS Trainium chips, with minimal or no code changes for code already running on other accelerators. It adds an optional "adaptive eager" mode that groups and compiles operations, plus multi-level caching; a shared compilation service cut time to first training step by about a third in a 64-process test.

Oct 13, 2026Tue
  1. Amazon Science75

    AWS priming cuts hybrid model pretraining cost by over 99%

    Amazon Web Services researchers introduced priming, a method that reuses pretrained Transformer weights to build hybrid state space models, cutting pretraining cost by more than 99% and using under 0.5% of the source model's token budget. A 32-billion-parameter Gated KalmaNet hybrid matched its source Transformer on reasoning while decoding up to twice as fast at 128,000-token context. Code and models were released under Apache 2.0.

    Why it matters: Cheap, controlled architecture experiments could accelerate the shift away from pure Transformers for long-context agents, where KV-cache memory costs dominate serving economics.