#Data/Training

AIDatabricks introduced Vibe Data Modeling, an open-source agent that helps teams build, validate, and evolve business-specific data models. It applies roughly 250 modeling rules while keeping data modelers and business stakeholders involved. Teams can start from 40 industry models as a baseline and iterate toward models that reflect how their business operates.

AIActor Ben Affleck drew attention this week for explaining machine learning concepts, including convolutional neural networks, tensors, and transformers, in several recent interviews. He said he fine-tuned open video models by unfreezing weights and training only the last cinematic layer, using a dataset he built over about eight months for his startup. Affleck said he worries about students and learned helplessness more than Skynet, and predicted AI will be additive to the movie business.

AIGoogle Cloud's borderless Lakehouse lets Gemini query data on AWS and Azure without variable egress fees. It reads directly from Salesforce Data 360, SAP, ServiceNow, and Workday without copying data. It also federates open Apache Iceberg tables across Databricks Unity, Snowflake Horizon, and AWS Glue.

AIGoogle Cloud's Smart Storage enriches files in place and writes metadata directly onto source objects, giving Gemini instant context. The post says this keeps security ACLs from drifting, which matters for dark, unstructured data.

AILlamaIndex says Markdown has become a universal representation between humans and agents, preserving headings, lists, and tables while remaining readable to models. Since most unstructured documents are not natively in Markdown, the main challenge is the translation layer, which the company addresses with models that convert document containers into Markdown. The quoted post adds that Markdown keeps table columns intact, with HTML used for tables with merged headers.

AIOpenDocRouter is a unified API and billing interface for document OCR models, ranging from lightweight open-source options like MinerU to frontier VLMs like Opus 5.5. Per the linked post, models are served at cost with a small transaction cut, rate limits are handled, and bounding boxes and layout are offered as a service.

AI推荐阅读！ ExploreNet: Learning Where to Explore in Diffusion GRPO ExploreNet is a method introduced by @StellaLisy for learnable exploration in diffusion RL. It learns an adaptive exploration distribution conditioned on the current state, rewarded by the diversity of rollouts, aiming for faster, more targeted learning in diffusion GRPO.

AIRSIGym provides a research agent with training, inference, evals, and sandboxes as callable services, so it spends its budget on experiments rather than rebuilding infrastructure. With Opus 5 as the researcher, the improved system rose from 17.67% to 50.33% on SWE-bench Verified. The post also highlights a way to measure co-evolution between harnesses and models.

AIOdyssey announced Odyssey-3, a foundation world model it says enables applications in physical AI, human experiences, and training intelligences. The company highlights agents learning from experience inside Odyssey-3 while working toward objectives.

AILlamaIndex says Markdown is its default output for document parsing because it preserves headings, lists, and tables, which helps models read content correctly. The post notes that parsers can extract every word yet lose which column a number belongs to, forcing models to guess. For tables with merged headers, LlamaIndex switches to HTML.

AIPeriodic Labs' Liam Fedus and Ekin Dogus Cubuk explain why scientific discovery differs from math and coding, and why experiments remain the ground truth. They describe reinforcement learning grounded in physical experiments, AI-driven materials characterization, and the view that failed experiments can be valuable training data. The transcript was truncated before the discussion of giving lab instruments "140 IQ" was completed.

AIGoogle Research is presenting EnvHarness, a flexible framework that enables co-evolution between LLM agents and their training environments, at the #COLM2026 Google booth #107 today at 11:00 AM PT. The post notes that static environments limit agent growth, and EnvHarness is described as a plug-in architecture that dynamically reshapes environment behaviors to improve reinforcement learning and adaptability.

AIVoice Arena's Monsoon ASR dataset fine-tuned Whisper Medium on Bengali FLEURS, reducing LLM word error rate from 85.27% to 7.65%. The corpus spans 100,000 hours across 50 languages, and Voice Arena says more than 80 organisations have asked to license it since its launch a week ago.

AIFireworks AI argues that serious AI teams eventually build their own data, evaluations, models, and possibly chips. The post features swyx of Latent Space and AI Engineer discussing why the future is domain-specific, and promotes an in-person event in San Francisco on November 3.

AIIn a prospective clinical study, Google's models pinpointed gestational age within 4 days of accuracy. The company says that precision could meaningfully affect clinical care, and that extending such tools to low-resource settings could help reduce maternal deaths and close care gaps worldwide.

AICarbon-A is an open model that predicts gene locations directly from DNA, and it has been used to annotate genomes from over 22,000 species. The release includes a database of 566 million gene candidates, about 16 times the gene annotations in the RefSeq dataset. Wet-lab RNA experiments supported 239 candidates missing from RefSeq across cats, Syrian hamsters, chickens, and Arabidopsis.

Why it matters: The source ties an open gene-annotation model to specific wet-lab checks and gene counts, helping readers judge how far its predictions extend beyond well-studied genomes.

AIThe team released Carbon-A, an open model that finds genes directly in DNA, along with a database of 566.34 million candidate genes across 22,617 species. The model reads genomes without needing a close relative, and wet-lab validation in cats, chickens, and arabidopsis is cited, with 239 genes found missing from reference annotations of common species. The authors say the model marks gene locations but does not design DNA or predict gene function.

AIZyphra has released a blog post and arXiv paper on expert coupling in mixture-of-experts (MoE) pretraining. The post itself provides only links, so the specific findings are not described here.

AIZyphra says its approach is lossless: the same tokens still reach the same experts, with unchanged architecture, routing decisions, and training objective. By reorganizing where experts and tokens live, it reduces the communication needed to perform the same computation.

AIZyphra reports that its MoE training optimizations speed up token exchange by 1.16x to 2.63x and full training steps by up to 1.41x in Megatron-LM on 8 to 64 GPUs. The gains are largest when each token uses more experts and those experts span several nodes.

AIZyphra reports that patterns in model training emerge early enough to be useful and are inexpensive to measure. Using only 4,096 tokens, its placement and routing predictions come within roughly one percentage point of results obtained with 524,000 tokens.

AIZyphra reports that expert routing in mixture-of-experts models is predictable across layers, since the experts a token uses in one layer indicate which it will need next. Its token shuffling method moves each token to the GPU holding those experts within a transfer that already runs after attention, adding no network traffic.

AIZyphra places experts that are frequently chosen together on the same GPU, so each token is sent to a GPU only once regardless of how many of its experts reside there. On 8 GPUs, this removes up to 58% of token copies sent across the network.

AIIn Zyphra's runs, MoE token exchange between experts consumed 13-24% of step time on one node and 45-60% across four nodes. Because experts are spread across GPUs and nodes, tokens must be sent to their experts and returned, making this communication a major training cost as models scale.

AIIn one layer, just 64 of 8,128 possible expert pairs handled 42% of tokens, far above the 1.6% expected if experts were chosen independently. Zyphra says this structure can be exploited to improve expert placement.

AIZyphra researchers report that exploiting patterns in how tokens are routed to experts makes Mixture of Experts communication up to 2.63x faster on AMD MI300X GPUs. The speedup comes without changing the model itself.

AILarge-scale AI training is spreading across multiple datacenters, with Google, Microsoft, AWS, Meta, and CoreWeave cited as examples. Because synchronized GPU clusters must exchange data in bursts, inter-site links can become a bottleneck, which Cisco estimates may require aggregate bandwidth about 14x a conventional DCI baseline.

AIThe author argues that agents working across a company fail because they lack the decisions and context recorded in threads, meetings, and DMs, not because the model is weak. The approach stores distilled claims with source evidence and time, never overwrites facts, labels missing information explicitly, and resolves permissions before the model runs. The report cites results on LongMemEval, including 99.8% top-ten evidence recall and $8.24 ingestion cost, and says an open-weight model can match frontier extraction quality.

AINVIDIA announced commitments valued at $1 billion over the next five years to build U.S. capacity for super intelligence research in fields including quantum computing, healthcare and energy security. The funding will support U.S. higher-education research institutions, American quantum leadership and cloud service providers serving U.S. government mission needs. NVIDIA is also a collaborator on several phase 2 Genesis Mission awards in quantum computing, fusion, accelerator design and microelectronics.

AIAtomic Canyon's Nuclear Industry Virtual Assistant (NIVA), developed with nuclear-industry groups, is now available to the entire U.S. fleet of 94 reactors after pilot testing at Constellation Energy plants. Nuclearn says its products have reached more than 65 U.S. partners, and the article says the industry is turning to AI to help manage regulatory paperwork and a shrinking, aging workforce.

AIGoogle has announced a single, universal Gemini agent for Gemini Enterprise as part of its Gemini at Work updates. The agent answers questions, handles knowledge work, creates images and media, and writes and runs code. It works inline in Gmail, Drive, Docs, Slides, Sheets, Chat, and Calendar, with new data and analytics skills for plain-language insights and industry-specific tools for financial services and legal teams.

AIReJev, an independent community project, applied LoRA post-training to OpenBMB's MiniCPM5-2B for bounded agent decisions: state, question, and candidate options yield one choice. On its sealed 1,892-sample holdout, accuracy rose from 51.11% to 80.50% (+29.39 percentage points) with 0% invalid outputs, at about $5.31 in cumulative Modal billing including earlier experimental overhead. The authors describe this as an early, task-specific result, not parity with Jev.

AIAmazon Web Services launched an open-source Physical AI Toolchain that combines AWS services with NVIDIA's Physical AI software to cover data generation, model training, simulation, edge deployment, and continuous improvement for robots. AWS uses Amazon SageMaker for training and AWS IoT Greengrass for distributing models to edge devices, while NVIDIA contributes Isaac Sim, Isaac Lab, Isaac GR00T, and Cosmos. The toolchain is hardware-neutral and does not directly replace RoboMaker, which was shut down in 2025.

AIReplit and Databricks integration, now generally available with native Lakebase support, lets enterprise teams build apps from plain-language prompts using Replit Agent and deploy them as Databricks Apps. Deployed apps inherit automatic user authentication and Unity Catalog access controls, and Replit Agent auto-provisions a managed Lakebase Postgres database for operational data. Lakebase keeps app-written data inside the Databricks perimeter instead of a separate external database.

AIIBM's torch-spyre integration makes Spyre, its dataflow inference accelerator, a native PyTorch device by mapping PyTorch's device, allocator, stream, and event abstractions onto the Spyre runtime and firmware. Tensors stay resident on device="spyre" between operations, and FX graphs remain in the Inductor compiler path. The approach gives eager and compiled execution one path with lower launch overhead.

AIA new poll finds 64% of Americans believe AI is developing "too fast." The source text provides no further details on the survey's methodology, sample, or other findings.

AIEthan Mollick, who ran early RCTs on AI chatbot productivity, says few similar studies have appeared since true AI agents emerged last fall. He attributes this partly to newness and partly to research design challenges, and suspects large effects may be going unmeasured.

AIElevenLabs explains how to build meeting transcription products using its Scribe v2 and Scribe v2 Realtime models through its API. Real-time transcription suits live captions and in-meeting bots, while batch transcription suits post-meeting notes and records, with Scribe v2 Realtime reporting 150 ms latency and supporting up to 50 key terms for prompting.

AITian Keyu's unnamed startup has raised nearly $30M from 5Y Capital and IDG at a $200M post-money valuation, according to Bloomberg. The NeurIPS 2024 award-winning researcher's 10-person team is developing a 200,000-symbol visual vocabulary to help AI process video. Tian claims the approach could cut video-generation costs at least tenfold, with a full model release planned for 2027 and no product yet.

AIA HP survey of nearly 20,000 desk-based workers globally found three in five UK business leaders say AI adoption has created new roles or teams in their organizations. Only 10% said AI is primarily replacing or reducing roles, while 52% of UK workers use employer-provided AI tools daily or weekly, up from 38% last year. Still, 38% of UK knowledge workers worry AI could replace their roles.