Skip to content

#数据/训练

Today 1 items
TodayOct 18Sun
  1. TechCrunch · AI55

    Meta removes language that raised scraping and AI training concerns

    Meta removed wording in its commercial terms that allowed content from business-account websites to be used to develop, improve and train its products. Legal experts had previously said the clause could authorize Meta to scrape any publisher's website with a business account to train AI models. Meta responded that the consolidated scraping clause applies only to ads and commercial products, and that its use of web crawlers has not changed.

Oct 17Sat
  1. The Verge · AI57

    Meta removes language that may have allowed it to scrape news sites

    Meta updated its commercial terms, removing wording interpreted as allowing scraping of content including news sites for AI training, which Cornell digital and information law professor James Grimmelmann and others said could override robots.txt blocking directives. Meta spokesperson Daniel Roberts responded that the scraping clause applies only to ads and commercial products, the company's use of web crawlers has not changed, and content owners can still express preferences via robots.txt.

Oct 16Fri
  2. Hugging Face Blog50

    Scaling Indic Pretraining: 2.8 Trillion Tokens and an Open Data Generation Pipeline

    NVIDIA and IIT Madras's AI4Bharat released the Nemotron-pretraining-Indic-v1 pretraining corpus, roughly 2.8 trillion tokens covering all 22 scheduled languages of India, with 2,601.0B natural-script tokens and 224.6B romanized tokens (Gemma-4 tokenizer statistics); it will soon ship alongside an open-source, GPU-accelerated data generation pipeline.

Oct 15Thu
  1. Semiconductor Engineering Chips and Packaging46

    Characterization of HBM Power Consumption (ETH Zurich, NYU)

    Researchers from ETH Zurich and NYU published a paper presenting the first comprehensive experimental characterization of power consumption on real HBM2 DRAM chips, with 13 new findings and 6 key conclusions. Read operations draw significantly more current than other operations, and flipping all bank group bus signals increases power by 29.4% over holding them constant. Based on this, the team constructed access patterns that on average consume 23.6% more power than the JEDEC-specified maximum power measurement loop across 36 tested HBM2 chips.

Oct 14Wed
Oct 13Tue
  1. MIT Technology Review · AI69

    AI labs are buying data from dead startups to train their models

    MIT Technology Review reports that as public and synthetic datasets peak, AI labs are buying code and workspace data from dead startups via shutdown service providers such as SimpleClosure and Sunset. SimpleClosure's Asset Hub has facilitated over 150 transactions and is expected to recover over $10 million for startups by the end of 2026; 47% of its transactions now include code or data sales.

  3. Microsoft Research40

    Simulating student learning for better AI tutoring

    Microsoft Research released StudentSim, a research system for evaluating and improving AI tutoring tools before testing with real learners, training a personalized simulator for each student based on publicly available de-identified learner records. It also introduces the StudentSimEval benchmark, which evaluates along two dimensions, behavioral fidelity and instructional responsiveness, across three domains: chess, second-language English writing, and math, outperforming leading baselines. The system is for research-stage use only, is not a product, has not been deployed in classrooms, and is not intended to replace teachers.

Oct 12Mon
Oct 9Fri
Oct 8Thu
Oct 7Wed
  2. Hugging Face Blog43

    Falcon-Emirati: When an LLM Learns the Dialect, the Culture, and the Nuance

    The co-publishers with Hugging Face released Falcon-Emirati-7B, fine-tuned from the Falcon-H1-Arabic 7B variant to understand and generate Emirati Arabic dialect. The model uses a hybrid Mamba and Transformer attention architecture, trained on Emirati dialect web data, Emirati-culture MSA data, and synthetic data constrained by lexicon and style rules, with a context window supporting up to 128K/256K tokens.

Oct 2Fri
  2. Google Research51

    Toward provably private learning from federated data

    Google has announced a next-generation federated learning system based on trusted execution environments (TEEs), providing remotely verifiable and auditable guarantees for data anonymization. Access policies are published to the public transparency log Rekor, the KMS and data processing binaries are reproducibly built from open source code, and encrypted training data is decrypted and processed only inside the TEE, with workload operators seeing only metrics and differentially private model weights.

  3. Hugging Face Blog39

    AutoSynthData: Generating Training Data for Enterprise Agents

    ServiceNow CoreAI launched AutoSynthData, which uses a target model's failures in an environment and successful trajectories from stronger teachers to automatically generate executable, realistic, and hard training tasks for enterprise agent post-training. Each task consists of a system specification, user prompt, and verifier, and must meet feasibility, realism, and difficulty requirements. The team demonstrated the pipeline with a public dataset from EnterpriseOps Gym and dynamically adjusted the curriculum as models improved.

Oct 1Thu
  2. The Decoder78

    Meta dodges billions in US taxes by calling its AI data centers experiments

    Meta classified its AI data centers as "experimental models" and its Nvidia chips as experimental materials, using the federal R&D tax credit to save $3.9 billion in 2025, up from $2 billion the prior year and $700 million in 2023, making it the largest beneficiary of the credit among public companies.

    Why it matters: The piece gives specific tax-savings figures over time and SEC risk disclosures, letting readers see how AI infrastructure investment is monetized through the 1981 R&D tax credit.

Sep 28Mon
Sep 26Sat
  1. AWS Machine Learning Blog31

    Scaling MoE reinforcement learning on Amazon EKS with EFA and DeepEP with 40% more throughput

    AWS proposed an architecture based on Amazon EKS, EFA and DeepEP to accelerate large-scale reinforcement learning post-training of MoE models, delivering 40% higher throughput. The solution targets workloads like RLHF, PPO and GRPO, addressing resource balancing between rollout generation and policy training, high-throughput communication across hundreds of accelerators, and the dynamic all-to-all token routing overhead introduced by Expert Parallelism.

Sep 24Thu
Sep 22Tue
Sep 16Wed
Sep 11Fri
Sep 10Thu
Sep 3Thu
  1. Hugging Face Blog62

    Training a coding model to paint watercolours with TRL and OpenEnv

    A Hugging Face blog author used TRL and OpenEnv to openly reproduce Surya Narreddi's approach, training Qwen/Qwen3.5-35B-A3B to write p5.brush code that paints watercolours.

    Why it matters: The author has open-sourced the training recipe, environments, reference pool, and model, so readers can reproduce the entire pipeline with a single command and learn from the troubleshooting notes.

Aug 28Fri
  1. Google Research50

    Planetary prediction engine: Automating global models via Earth AI

    Google Earth AI has introduced an experimental research capability called the Planetary Prediction Engine (PPE), which autonomously performs the full geospatial prediction workflow—data discovery, cleaning, feature engineering, model training, and evaluation—starting from natural language queries, compressing work that previously took weeks of manual data engineering down to a few minutes.

Aug 27Thu
  1. Google Research47

    GlucoFM: Foundation model for continuous glucose monitoring

    Google unveils GlucoFM, a self-supervised foundation model with a two-stream design that separates slow glucose baseline trends from short-term deviations. Across four cohorts, seven clinical prediction tasks and 14 evaluations, its mean PR-AUC improves from 54.7 to 58.8 over the strongest baseline CGM baseline, leading on all diabetes risk and beta cell dysfunction evaluations, and achieves the lowest MAE in postprandial glucose response prediction.

Aug 25Tue
  1. Hugging Face Blog61

    Granite 4.2 LLMs: How They're Built

    IBM released the Granite 4.2 reasoning model family, comprising 3B, 8B and 30B dense decoder models built on the Granite 4.1 base (pretrained from scratch on about 15T tokens, with a five-stage strategy extending context to 512K), post-trained with SFT and multi-stage GRPO reinforcement learning, all open-sourced under the Apache 2.0 license.

    Why it matters: The official disclosure covers the full training pipeline from pretraining and SFT to multi-stage GRPO and agentic RL, transferable to similar model-building workflows.

Aug 10Mon
Jul 26Sun
  1. Berkeley AI Research46

    Teaching LLMs to Update Beliefs for Efficient Long-Horizon Interaction

    Berkeley AI Research proposes the ABBEL framework, which replaces full interaction history with a natural-language belief state as the agent's working context and introduces belief grading to supervise summary content. On the CollabBench collaborative coding benchmark, ABBEL-rec-BG closes roughly 50% of the performance gap with full-context models, cuts training steps from 100 to 50, and uses significantly shorter peak context token lengths than the full-context setting.

Jul 9Thu
Jul 6Mon
  1. Hugging Face Blog47

    PRX Part 4: Our Data Strategy

    Hugging Face details the data pipeline for its PRX image generation model: mixing public and internal datasets for pretraining, using a VLM to regenerate long captions, then converting them into streamable training data. The text encoder has been switched to Qwen3-VL, with text latents computed online during training at a throughput cost of only about 3–4%. Images are uniformly encoded as JPEG quality 92, and the data is built with Lance and streamed for training via MDS (Mosaic Streaming).