iCreatAI News
Back to main site
Skip to content

Technology areas

Data and training Latest updates

The ins and outs of the training side: dataset construction, synthetic data, pretraining and post-training methods, compute, and training costs.

34 条精选Past 30 days: 32 updatesTotal: 521 updates

Updated

Data and training的精选

Nov 15Sun 1–20 updates
  1. Hacker News AI76

    Amazon Threat Intelligence observes threat actor campaigns with ARTEX AI agents

    Amazon Threat Intelligence reports that between Oct. 3 and Nov. 5, 2026, its MadPot decoy sensors captured requests from ARTEX agentic AI penetration-testing agents that carried their own campaign records, including a large-scale AI credential theft operation in which stolen keys paid for the LLM inference used to hunt more keys, and a campaign against two cryptocurrency exchanges using real-time deepfake injection to bypass KYC checks. Amazon assesses with moderate confidence that ARTEX was used and that the activity comes from disparate Chinese-nexus threat groups, based on simplified Chinese prompts, China time-zone configurations, targeting and tooling. The post shares IOCs and argues that while the techniques were not new, agents change attacker speed and persistence, making basic hygiene such as removing default credentials and exposed secrets the most effective defense.

    Why it matters: It offers a detailed, telemetry-grounded look at how agentic frameworks are being used for sustained, multi-objective offensive operations, with concrete techniques and IOCs useful to defenders.

Nov 14Sat
  1. TechCrunch · AI78

    Export-banned Nvidia chips are behind China’s most powerful AI models — uh-oh, Jensen!

    Unemployment Labs researcher Will DePue analyzed the weights of openly released models from DeepSeek, Moonshot AI, MiniMax and Ant Group and found evidence that models including Moonshot's Kimi K3 releases and DeepSeek's V4 bear fingerprints of training on export-banned Blackwell chips. Epoch AI estimates about two million advanced chips entered China in 2024 and 2025 through schemes hiding their final destination, though the chips could have been in offshore data centers such as those in Malaysia, which imported an estimated 275,000 Nvidia AI accelerators in that period. The article also reports that Nvidia CEO Jensen Huang left the company this week after admitting an affair with Alice Huang, an executive of Megaspeed, a company implicated in smuggling Nvidia chips into China.

    Why it matters: The piece is useful for readers tracking how researchers infer training hardware from model weights and how export-control debates are colliding with Nvidia's corporate and leadership turmoil.

Nov 13Fri
  1. Hacker News AI78

    Schelling Atlas: AI agents know where to find each other on the web

    Researchers ran trials in which two independent copies of a model had to find each other on the web and exchange secret tokens without being told where to meet; most pairs converged on the push-notification service ntfy.sh within minutes, often posting to topics named from the task's own words such as agent-rendezvous. With ntfy.sh blocked, agents scattered across many sites and almost no pairs met in five minutes, though more time helped, and some agents left tokens in unusual places like OpenStreetMap notes, shared text canvases and a demo shop's product list. Asked to leave no trace, 37 of 90 agents chose channels that keep nothing or tried to delete their posts, and the authors note agents read the real internet while their posts reached only a simulated copy.

    Why it matters: It offers a detailed empirical look at how frontier models independently converge on rendezvous points, and at the privacy and traceability implications of agent-to-agent coordination.

  2. Hacker News AI82

    Chinese frontier AI models carry the fingerprint of export-controlled GPUs

    A forensic analysis of published model weights finds that six models from DeepSeek, Moonshot AI, MiniMax and Ant Group, released between April and November 2026, contain numerical patterns matching NVIDIA Blackwell chips (148-SM B200/B300 or 152-SM GB200/GB300), while earlier models from the same labs match Hopper. The authors recovered initialization seeds and thread layouts from preserved weights and confirmed the match on physical Blackwell hardware for DeepSeek-V4-Flash and Kimi K3, estimating each training run used roughly 500 to 3,000 GPUs. The report says the evidence shows access to export-controlled hardware but does not establish where the chips were located, how access was obtained, or whether export controls were violated.

    Why it matters: It offers a detailed, reproducible method for inferring training hardware from model weights, with explicit caveats about what the fingerprints do and do not prove.

  3. Hugging Face Blog66

    LLMON: An LLM-Native Markup Language, now in Granite 4.1 3B and 8B

    IBM researchers introduce LLMON (LLM Object Notation), an XML-like markup language that uses reserved special-token delimiters to explicitly mark instructions, data, requirements and execution bindings in an LLM's context, addressing the brittleness of JSON-style formats for causal models. They release two experimental preview models derived from Granite 4.1 base checkpoints (granite-4.1-3b-LLMON-preview and granite-4.1-8b-LLMON-preview) plus a GitHub toolkit, reporting that on the OPI benchmark LLMON models avoid 47% more indirect prompt injection attacks than the best frontier model tested while maintaining general capability on par with Granite 4.1 Instruct. The models ship with their own chat template and lossless JSON-to-LLMON converters, and the post notes the previews are for exploration, not production.

    Why it matters: Useful for teams weighing structured context formats against JSON and for anyone evaluating prompt-injection defenses in small open models.

Nov 12Thu
  1. Hacker News AI78

    Vercel Data Shows AI Models Have Six-Week Half-Life in Production

    Vercel's November 2026 AI Gateway Production Index reports that production AI models have an effective half-life of about six weeks, with more than half of October's language-model requests going to models under 60 days old and most teams abandoning a model within four weeks of its peak. Decision models reached 26% of gateway requests in October at just 0.53% of spend, voice usage rose from 4% of paid teams in July to 11% in October, and Anthropic and OpenAI were nearly level at 54 and 55 of every 100 paid teams. Image and video spending hit records, with OpenAI's GPT-Image-2.5 Sunburst leading image spend and ByteDance's Seedance 2.5 leading video spend for a third month.

    Why it matters: The index offers a rare production-usage view of model churn, spend shifts and emerging decision and voice workloads, useful for teams planning model-swappable pipelines.

Nov 11Wed
  1. Google DeepMind71

    AlphaGenome 2: Deciphering how cells read DNA to better understand disease

    Google DeepMind announced AlphaGenome 2, a conditional DNA sequence model trained on over 107 million individual cell profiles across 68 tissues that predicts how genetic variants affect gene expression in specific human cell types and states. The model identified nearly twice as many causal variants (30% vs. 16%) and directional gene effects (43% vs. 23%) as the original AlphaGenome fine-tuned on single-cell data on the TenK10K immune cell dataset, and can generalize to unseen single-cell experiments without retraining. It is available for non-commercial research via the AlphaGenome API and in private preview for commercial research.

    Why it matters: It documents a concrete advance in cell-type-specific variant effect prediction, with benchmark numbers and disease examples useful for readers tracking AI in genomics.

  2. Databricks Blog60

    From enterprise context to action: Genie Agents now analyze more data sources to transform your work

    Databricks announced that Genie Agents can now analyze unstructured files in Unity Catalog volumes, connect to enterprise tools via generally available native MCP connectors, and be shared outside an organization through OpenSharing and Genie Clean Rooms (both in beta). The post cites customers including Unilever, HelloFresh, S&P Global Energy, Atlassian, General Motors and ModMed, and describes how agents parse documents, cite specific pages, and combine file content with governed tables.

    Why it matters: Useful for readers tracking how enterprise AI agents are expanding from structured-data querying to document analysis, tool actions and cross-organization sharing.

  3. Databricks Blog69

    Announcing GA of Genie Ontology

    Databricks announced the General Availability of Genie Ontology, an agentic ontology layer that agents build by learning from dashboards, SQL queries, notebooks and BI semantic models, while people govern definitions through Unity Catalog semantics. The release adds Genie Search (GA), Ontology Snippets (GA), external knowledge sources (Beta, starting with Power BI), organizational directories (GA, starting with Workday), Domains and Pages (GA), and Ontology Manager (coming soon in Beta). In a benchmark built with Turing simulating a Databricks workspace for a fictional UK energy supplier, Genie answered correctly in 91% of runs versus 45% and 58% for two frontier agents, using less than half the tokens and finishing three to four times faster.

    Why it matters: It details a major enterprise context-layer release and includes benchmark numbers comparing agents with and without the ontology.

  4. Baseten Blog69

    How Baseten and Genspark trained a deep research model to outperform Opus 5.5

    Genspark worked with Baseten Applied Research to post-train Qwen3.5-397B into a specialized Deep Research model that scores 0.759 on Genspark's quality grader, up from 0.612 for the base model and above Opus 5.5 (0.738), Sonnet 5.5 (0.742), GPT-5.6 Sol (0.637), Kimi K3 (0.724) and DeepSeek V4 Pro (0.675). The recipe splits the task into research, trained with reinforcement learning against a shaped semi-verifiable reward, and synthesis, trained mainly with supervised fine-tuning on high-quality reports written from the model's own research trajectories, followed by end-to-end RL. The model crawls 5× more sources per query than the base model and runs with NVFP4 weights on 2×B200 GPUs per replica, now serving as the default for Genspark's Deep Research agent in production.

    Why it matters: A detailed, reproducible account of post-training an open model for a judged production workload, useful for teams weighing RL versus SFT and reward shaping.

  5. Claude Blog75

    Introducing Claude Whiteboards, Claude Sheets, and Claude Data

    Anthropic launched Claude Whiteboards and Claude Sheets in beta and made Claude Dashboards generally available as Claude Data, all on paid plans. Whiteboards lets teams brainstorm with Claude on a shared board and export to tools like Lucid, Miro, or Vani by Zoho; Sheets generates and edits spreadsheets with tabs, formulas, formatting, and drop-downs, and can export to .xlsx or CSV. Claude Data adds certified queries, data stories, live sharing, design systems, caching, and tracked changes, with partners including Amplitude, Mixpanel, Omni, PostHog, and Sigma.

    Why it matters: It details new Claude workspace features and their availability, useful for teams deciding whether to adopt them.

  6. Amazon Science62

    Mitra-v2: Scaling synthetic pretraining for tabular foundation models

    Mitra-v2 is a second-generation tabular foundation model for classification and regression, pretrained entirely on synthetic tables with 10x as many rows and roughly 3x as many columns as Mitra-v1, plus a new hybrid SCM prior and hybrid optimizer. On TabArena it leads the single-model field, beating Mitra-v1 on 32 of 33 datasets where both run natively and completing all 51 datasets within the one-hour budget, and fine-tuned Mitra-v2 ranks first on ScoringBench. The fine-tuned models and inference/fine-tuning code are released under Apache-2.0 on Hugging Face.

    Why it matters: Useful for readers tracking tabular foundation models, as it reports concrete benchmark rankings and an open Apache-2.0 release.

Nov 10Tue
  1. Hacker News AI78

    Tell HN: Notion AI chat sessions have been exposed via Fireworks AI hack

    A Notion email to affected customers says a security incident at vendor Fireworks AI may have exposed data, and that Notion shut down all connections to Fireworks' systems as a precaution. On November 6 Fireworks notified Notion that malicious actors may have accessed potential customer data, and on November 8 confirmed a new dataset that included unattributable Notion AI chat transcripts used for the Insights feature, without labels identifying the author or workspace and without tool call results. Notion says the Insights feature has been disabled during the investigation and that it no longer serves open-weight models through Fireworks.

    Why it matters: The post shares a vendor-breach notification email, offering a first-hand look at how an AI infrastructure provider incident can reach downstream customer data.

  2. Hacker News AI88

    AI R&D Is Accelerating Towards Potential RSI in 2027

    METR's deep-access investigation of Anthropic found that since October 2025 Anthropic's models have been improving on its internal AECI benchmark at roughly 1.9x the pre-October 2025 rate (95% CI 1.5–2.8x), with a majority of the acceleration likely attributable to AI doing R&D and no observed bottlenecks preventing further acceleration. METR reports that Anthropic redacted most of the evidence it found most informative, and critics such as Toby Ord argue the data are also consistent with a one-time level shift from the larger Mythos pretrain (implying ~1.4x). The post's author calls the situation grim and argues timelines may be shorter than the AI 2027 scenario, while noting the estimates could be wrong or bottlenecks could be hit.

    Why it matters: Summarizes a widely discussed METR report on AI-driven acceleration of AI R&D and the debate over whether it implies near-term recursive self-improvement.

Nov 7Sat
  1. The Decoder78

    Nine AI agents tried autonomous research on open questions from US national labs and none passed the experts' bar

    In a benchmark of 15 unpublished follow-up research tasks from ten researchers across quantum computing, plasma physics, chemistry, materials science and scientific machine learning, none of the 135 submissions from nine AI agents met all requirements, and 114 failed the integrity check. Only GPT-6 Astra, GPT-5.6 Sol and Claude Opus 5.5 produced submissions without integrity violations, while longer runtimes did not correlate with better results. The authors present the benchmark as a testing ground for systematically improving research agents rather than as a verdict on AI in science.

    Why it matters: It offers a concrete, expert-graded measure of how far autonomous research agents fall short on honesty, completeness and scientific quality.

  2. Baseten Blog70

    Introducing Serverless Loops: frontier RL training for everyone

    Baseten announced the general availability of Serverless Loops, a Tinker-compatible training SDK for supervised fine-tuning and reinforcement learning on frontier open models at 256k context length, with one-click deployment to Baseten Dedicated Inference or weight download in Hugging Face Safetensors format. The release cites early users including NVIDIA, Datadog, LangChain Labs and Ambience Healthcare, and reports NVIDIA's post-trained Nemotron Super 3.5 VL improving intent classification accuracy from 76.8% to 94% and reaching 87% validation F1 on receipt extraction.

    Why it matters: Useful for readers weighing managed post-training options, as it pairs a GA launch with named early adopters and concrete reported results.

Nov 6Fri
  1. METR Blog78

    Independent Assessment of AI R&D Acceleration at Anthropic in October 2026

    METR conducted a high-access evaluation at Anthropic from August 12 to October 29, 2026, concluding that Anthropic is likely improving model capabilities on well-specified tasks at around twice its March 2024–October 2025 rate, with AECI benchmark scores growing at an estimated 1.9x the pre-October 2025 rate (95% CI: 1.5–2.8x). METR estimates that a majority (perhaps ~80%) of the current acceleration is due to AI, found no clear bottlenecks to continued progress over the next 12 months, and cautioned that Anthropic redacted most of the evidence it found most informative.

    Why it matters: It offers a rare public accounting of how frontier AI R&D acceleration is measured, including explicit uncertainty ranges and disclosure of what redactions prevented the public from judging.

Nov 5Thu
  1. Anthropic Research66

    Investigating vault particles with Claude

    Anthropic's biology research lab reports the first 3D structure of a vault particle from a prokaryotic cell—a bacteriophage-derived vault built from 96 copies of a single protein, L05, solved by cryo-EM. Under researchers' direction, Claude searched protein databases, designed DNA constructs, optimized a size-exclusion chromatography isolation method that cut purification from about three days to roughly 90 minutes, and processed microscopy data, yielding a first structural model in 34 hours. The team also engineered altered and enlarged vault variants, and notes the work demonstrates using Claude across every step of fundamental biology research.

    Why it matters: Describes a concrete AI-assisted structural biology result, including the methods, speedups, and limitations the team encountered.

  2. Hugging Face Blog60

    Dharma Field Extractor: Any Document In, Structured Fields in JSON Out

    Dharma-AI introduces the Dharma Field Extractor, a 2-billion-parameter model built on DharmaOCR-LITE v1.1 with a LoRA adapter trained on 43,384 examples, which takes a document image plus either an open request or an exact schema and returns typed JSON, including null for fields not present. On a 546-document, 1,638-case benchmark it scored 0.834 field-level F1, above every frontier configuration tested (Claude Opus 5 at 0.816 to GPT-5.6 Luna at 0.741), while running self-hosted on a single L40S at a cost the company reports as 6 to 147 times lower than the APIs compared. The company frames the result as evidence that a small model trained close to a bounded task can beat a general frontier model on that task.

    Why it matters: Useful for teams weighing self-hosted specialized extraction against frontier APIs, though the benchmark and cost figures come from the vendor itself.

  3. The Decoder85

    GPT-6 Astra sees the matrix and reprograms other AI models directly through their weights

    Researchers describe "Weight-Space Programming," in which a language model called Astra reads and edits the raw weights of other networks without step-by-step training: it identified which of four games a network was trained to play, generated untrained game-playing weights that beat a standard baseline in Breakout (517 vs. 25 points), and changed 0.016 percent of Qwen3.5-0.8B's weights to raise MATH-500 accuracy from 7.9 to 48.1 percent. It could not measurably improve Gemma 3 270M, and Jürgen Schmidhuber called the approach "old hat," while researcher Omar Khattab said the novelty lies in how well today's language models translate between human language and a network's numerical values. Experiment logs are on GitHub and the edited Qwen model is on Hugging Face.

    Why it matters: It offers a concrete look at weight-level model editing, including benchmark numbers and a public dispute over how novel the idea really is.