Skip to content
Thursday, 19 November 2026
Tenesys AI News
Subscribe

AI Research

Research papers, benchmarks, architectures, reasoning and training methods.

“API” — 42 articles ✕

NVIDIA and AI4Bharat Release 2.8 Trillion-Token Indic Pretraining Corpus and Open Data Pipeline

NVIDIA, working with IIT Madras's AI4Bharat lab, is releasing Nemotron-pretraining-Indic-v1, a roughly 2.8 trillion token synthetic pretraining corpus covering 22 Indian languages, along with the open GPU-accelerated pipeline used to generate it. The corpus includes native-script data for all 22 languages and romanized text for twelve, built by translating curated English source material (web text, math, code, Q&A) using NeMo Data Designer and quality-checked with an MQM-inspired rubric. NVIDIA already used a 10-billion-token Indic subset for continued pretraining of Nemotron 3.5 Super VL, which showed measurable gains on Indic benchmarks.

Hugging Face

Apple Shows Trillion-Parameter-Class LLMs Running Locally on Mac Studio Clusters with MLX

Apple's MLX team published benchmarks showing large open-weight models — including a near-trillion-parameter MoE model called Muse Spark Open — running on Mac Studio hardware, both on a single machine and distributed across up to four Mac Studios using a new RDMA-over-Thunderbolt driver. The new M5 Ultra chip, with dedicated GPU Neural Accelerators and 1.2 TB/s memory bandwidth, delivers up to 4.6x faster time-to-first-token and 1.6x faster generation speed versus the previous M3 Ultra. For models too large to fit on one machine, MLX now supports pipeline and tensor parallelization across multiple Macs connected via Thunderbolt RDMA, letting a four-Mac-Studio cluster generate over 36 tokens per second on an 816B-parameter model.

Apple Machine Learning
Research· Important

Microsoft Research debuts Magnetar, a zero-knowledge proving system for ordinary code

Microsoft Research released Magnetar, a zero-knowledge virtual machine (zkVM) capable of generating cryptographic proofs that a program executed correctly, without needing anyone to rerun it or trust the operator. Instead of forcing developers to rewrite applications as custom cryptographic circuits, Magnetar proves standard RISC-V machine code, so programs written in Rust, Go, or C can be compiled normally and proven as-is. Benchmarks across 15 programs show it processing 7.2 to 8.7 million instructions per second on a laptop, with proofs checkable in roughly 12 milliseconds, and its cryptographic security depends only on established hash functions like SHA-256 or SHA-3 rather than elliptic curves or lattice-based schemes.

Microsoft Research
Research· 🧪 Worth Testing

AWS open-sources MoE kernel optimizations for vLLM, cuts latency and boosts throughput on H200

AWS detailed three kernel-level optimizations for Mixture-of-Experts (MoE) model serving on Hopper GPUs, implemented in collaboration with the vLLM community and shipped in Amazon SageMaker AI's LMI container. The changes target the inefficiencies of fused MoE computation during low-batch decode, which previously left Tensor Cores underutilized. Combined, the optimizations deliver up to 10% higher throughput and up to 9% faster time-to-first-token (TTFT).

AWS Machine Learning
Research· Important

IBM shows open-model AI agents can cut chip-design search costs by over 11x

IBM Research used a coordinated multi-agent 'agent factory' to explore hardware design optimizations for a quantum error-correction decoder, comparing proprietary and open frontier models under the same 100-hypothesis search budget. The open-model configuration, run on-premises via IBM's llm-d distributed inference framework, achieved a 52x speedup over baseline for about $26 in inference cost, versus $299 for a proprietary-model configuration that reached a higher 204x speedup. The study highlights that cache-aware routing and efficient serving (98% prefix-cache hit rate, ~$0.11 per million tokens) made the open-model approach dramatically cheaper while still finding valid, compact hardware designs.

IBM Research
Research· 🧪 Worth Testing

Apple researchers propose self-deciding context compaction for long-running AI agents

Apple ML researchers published a NeurIPS paper introducing SELFCOMPACT, a scaffold that lets language model agents decide for themselves when and how to compact their own accumulated context, rather than relying on fixed-interval summarization triggered by a token count. It combines a compaction tool the model can call with a lightweight rubric telling it when to fire (a sub-task is resolved) and when to hold off (mid-derivation or stuck). Tested on six benchmarks and seven models, it matched or beat fixed-interval summarization while using far fewer tokens.

Apple Machine Learning

Apple Researchers Propose Turning Idle Mac GPUs Into a Shared Inference Cluster

A new Apple research paper introduces Minions, a system that pools unused GPU capacity across Apple Silicon Macs into a single distributed cloud for running machine learning inference. Instead of buying additional GPU hardware, teams could tap spare compute sitting idle on existing macOS devices. The authors say the design handles task scheduling, failure recovery, and efficient use of available resources, offering an alternative when dedicated GPU capacity is scarce or expensive.

Apple Machine Learning
Research· Important· 🧪 Worth Testing

Amazon's 'Priming' Turns Pretrained Transformers Into Efficient Hybrid Models for Under 1% of Training Cost

Amazon Web Services researchers published a method called 'priming' that converts pretrained Transformer models into hybrid architectures mixing attention and state space model (SSM) layers, reusing existing weights instead of retraining from scratch. The approach cuts pretraining costs by over 99% and lets a 32-billion-parameter hybrid model match its source Transformer's quality while decoding up to twice as fast. Amazon tested three SSM types and released an open-source model zoo, training code, and serving tools under Apache 2.0.

Amazon Science

NVIDIA and Princeton Cut Decoding Bottlenecks in Hybrid SSM Models

NVIDIA, working with Princeton's Dao AI Lab, built ReplaySSM, a new way to store state in hybrid models that mix state-space (SSM) layers with attention layers. Instead of rewriting the entire state on every token, the method keeps a periodic checkpoint plus a short list of recent inputs, splitting the job of producing output from the job of updating memory. The result is roughly half the memory traffic, decode kernels running up to 1.80x faster at scale, and speculative decoding throughput more than doubling on 120B-class models, all while keeping accuracy close to full precision. The method already ships inside TensorRT-LLM, vLLM, and FlashInfer, and is running in production for GLM-5.3-Flash and Kimi-K3.

NVIDIA Developer

NVIDIA Opens Vera Rubin Rack Architecture and Pushes Open Standards for AI Data Centers

At the OCP Global Summit, NVIDIA contributed its Vera Rubin rack and compute tray designs to the open MGX architecture and announced a series of partnerships covering AI factory power, cooling and networking standards. Partners include hyperscalers such as Meta, Google, Microsoft and AMD, plus Cisco and Visionbay. The announcements cover rack design specs (MGX 1.2), new open-source software for managing AI data centers (DSX OS components), grid resilience standards, an 800VDC power architecture, new Ethernet switching (Spectrum-6) and a new optical transceiver standard (USFP) with Cisco.

NVIDIA

NVIDIA Shows How to Make Large-Scale Model Pretraining Bitwise Deterministic at Low Cost

NVIDIA described a workflow built into Megatron Core that makes large-scale pretraining bitwise deterministic, meaning independent or checkpoint-resumed training runs follow the exact same numerical trajectory. Using tensor fingerprinting across training phases, modules, operations and kernels, engineers can trace nondeterminism down to a specific kernel and fix it. On a trillion-parameter Nemotron model running on 2,432 GPUs, optimizations brought the performance cost of enforcing determinism (the 'determinism tax') down to about 2%, while keeping bitwise-identical results over 800 training steps.

NVIDIA Developer

Apple Publishes NeurIPS Paper on Fast Image Generation That Keeps Exact Likelihood Math

Apple researchers, with collaborators from the University of Pennsylvania and UIUC, introduced a new generative modeling technique called Normalizing Trajectory Models (NTM) in a paper accepted at NeurIPS. Unlike typical diffusion models that rely on many small Gaussian denoising steps, NTM treats each step as a flexible, invertible transformation that preserves exact probability calculations — something most fast few-step generation methods give up when they rely on distillation or adversarial training. According to the paper, NTM can match or beat strong text-to-image baselines using only four sampling steps.

Apple Machine Learning

Apple Research Proposes Rubric-Based Method for Training Agents Across Multiple Environments

Apple researchers, with National University of Singapore collaborators, published a paper on RISED, a new training framework for LLM agents that must learn across many different interactive environments at once. Instead of relying only on scalar reward signals, RISED uses an LLM judge to tag each training rollout with rubrics describing its behaviour, then uses these rubric tags to select training data and to supply extra supervision. The authors report RISED achieves the highest average success rate across environments compared to existing approaches, ranking first or second in every tested environment.

Apple Machine Learning

Google publishes five real-world studies validating its geospatial health model PDFM

Google Research released results from five partner-led evaluations of its Population Dynamics Foundation Model (PDFM), a Google Earth AI system that turns search trends, mobility, built-environment density and environmental data into monthly location embeddings. Health institutions plugged these pre-trained embeddings directly into their own existing statistical and ML pipelines, without any task-specific fine-tuning, and tested them across five very different epidemiological problems. Each partner reported measurable gains over conventional data sources, arguing for PDFM as a plug-and-play substitute or supplement to slow, siloed health datasets.

Google Research

Google Research publishes workshop report on privacy and security challenges for autonomous AI agents

Google Research released a report from its Contextual Agent Privacy and Security (CAPS) workshop, co-authored with over 50 academic and industry researchers. The report argues that AI agents need to reason about context-specific social norms, not just static permissions, to act appropriately when handling personal data and taking actions on a user's behalf. It proposes a multi-layered approach spanning system-level sandboxing, model-level reasoning, user-centric controls, multi-agent guardrails, and ecosystem governance, grounded in the theory of Contextual Integrity.

Google Research

AWS describes graph-based agentic AI for network root cause analysis

AWS researchers describe a system that models telecom networks as a continuously updated graph ('digital twin') and uses AI agents to run a cascade of graph algorithms (decomposition, clustering, centrality ranking) to pinpoint root causes of failures. The approach was demonstrated with NTT DOCOMO at Mobile World Conference, reportedly cutting root cause analysis from hours to minutes on commercial networks. Agents also consult an incident knowledge base and runbooks, and can escalate autonomy from advisory to supervised to full remediation over time.

Amazon Science

Microsoft Research launches Quine, an AI world model built for biological discovery

Microsoft Research has unveiled Quine, a research system combining a multimodal 'world model' of biology with an interactive harness linking AI models, scientific literature, lab tools, and researchers. Built jointly across data types such as genomics, proteins, chemistry, RNA/cell state, and bioimaging, the model aims to predict how biological systems respond to interventions before costly lab experiments are run. Working with the Broad Institute of MIT and Harvard, Microsoft used Quine to rank thousands of compounds for their potential to shift pancreatic cancer cells between therapeutic states, and the top-ranked candidates were validated in wet-lab assays within a single weekend. Microsoft is now opening a 'Quine Fellows' program to give selected scientists early access, while cautioning that the system is experimental, research-only, and not intended for clinical use.

Microsoft Research

Google's TPU Team Reproduces Ai2's Olmo 3 7B Training Run in MaxText

A Google Cloud TPU engineering team worked with Ai2 to rebuild Olmo 3 7B's pre-training from scratch using MaxText, Google's JAX/XLA training framework, running on TPUs instead of Ai2's original PyTorch/GPU setup. They matched Ai2's published results not just on the training loss curve but on four independent held-out evaluation surfaces across the full ~5.93-trillion-token, 1.41-million-step stage-1 run plus the stage-2 annealing phase. Along the way they ported Olmo 3's unusual architecture (reordered-norm blocks, QK-norm, 3:1 sliding/global attention ratio) into MaxText and caught a data-loader bug that had been quietly inflating apparent performance through memorization rather than genuine learning.

Google Developers