#Data/Training

AITypeSafe CEO Jev argues AI can solve extremely hard problems yet still fail at basic automation, so his company builds reliable decision-making models inside software rather than chat interfaces. He says the company rejects public benchmarks and API-layer refusals, and that data and task fit matter more than brute-force compute. He also says System One Models could reshape coding agents and software, and that he would not pre-train a model from scratch even with $1 billion.

AIXiaomi MiMo announced open-source releases of Pro and Flash, the MiMo-V2.6-Distill-Qwen-9B model, a technical report, over 7K RL task environments, an end-to-end RL framework, and composable mini-harnesses. The attached table shows MiMo-V2.6-Distill-Qwen-9B after SFT and after RL compared with Qwen3.5-9B, with RL scores higher on most listed benchmarks, such as SWE-bench Verified at 66.2 versus 60.0.

Why it matters: The table compares a 9B distilled model against Qwen3.5-9B on coding, cyber, and agent benchmarks, showing how the reinforcement learning stage changes results.

AIAlberto's second chart compilation argues the AI boom shows bubble signals, covering concentration in the top 10 S&P 500 companies at 40%, record datacenter cancellations, and historically extreme investor leverage. The piece also tracks hyperscaler capex heading past $1 trillion by 2027 and contrasts AI token output with actual labor productivity gains.

AIAmazon Bio Discovery developed three AI approaches for antibody drug design: MochiBind for sequence-based affinity ranking, CA-MAP for developability prediction with batch effect correction, and an agent-guided design system. The agent-guided system produced 46 lab-validated hits against a novel cancer target.

AIMike Knoop notes that logprobs used to be widely exposed by LLM inference APIs and sees the market maturing so that parts of the LLM stack can be packaged in new, useful ways. He links this to Bryan Helmig's post on prompting with max_tokens: 1 plus logprobs for fast, parallel judgments, which Helmig says has a lot more depth than he expected.

AIApple has released LensVLM-9B on Hugging Face, a 9B-parameter Vision Language Model that scans compressed images of text and selectively expands relevant pages to their uncompressed form. The repository provides a demo script and supports compression settings of 5x, 10x, and 15x. Model files are under the Apple Machine Learning Research Model License, and the accompanying source code is distributed separately under the Apple Sample Code License.

AIThe article explains how Mixture of Experts models change inference by making prefill, midfill, decode attention, and decode experts distinct workloads. It describes how KV cache state, expert routing, and parallelism choices shape compute, memory, and network demands across an inference cluster.

AIAmazon Science describes three papers on AI for antibody discovery: MochiBind ranks antibody binding strength from sequence alone, CA-MAP predicts developability properties using batch-aware context, and an agent-guided pipeline designed nanobody binders against a novel cancer target. In the pipeline, 116 candidates survived lab screening, and 46 were identified as strong binders, which are being used to train the next design cycle.

Why it matters: The source reports the method, benchmark setup, and experimental validation in a single design workflow, showing how predictors, agents, and lab screening connect in antibody discovery.

AILlamaIndex has added confidence scores to its Extract product, giving accuracy estimates field by field for extracted data. Developers can use these scores to decide which results their apps accept automatically and which need human review. The feature is available on the Cost Effective, Agentic, and Agentic Plus plans.

AILMSYS Org shared a blog post about NVFP4 KV cache, a topic linked from its 2026-09-16 article. The post itself contains only a link, so no further technical details, figures, or results can be confirmed from this source.

AILMSYS Org says NVFP4 KV cache in SGLang fits about 1.78x more context into GPU memory and speeds long-context decoding by up to 78%. Built with Alibaba Qwen and NVIDIA for Blackwell, it stores KV at about 56% of FP8's per-token footprint, with decode throughput up 37%, 58%, and 78% at 32K, 160K, and 1M context. The post reports near-lossless accuracy versus FP8 on GPQA-Diamond and AIME 2025 using Qwen3.5-397B-A17B, and it can be enabled with --kv-cache-dtype nvfp4.

Why it matters: The post gives specific memory and throughput figures for NVFP4 KV cache in SGLang, showing how the format trades cache footprint against long-context decode speed.

AIMicrosoft Research published a Nature paper on RetroChimera, a predictive model designed to accelerate chemical synthesis. The post says custom-made molecules advance medicine, materials, and agriculture, but producing them remains slow and expensive. The model is meant to help researchers explore a wider range of molecules.

AIMicrosoft Research published RetroChimera, a retrosynthesis framework that combines the R-SMILES 2 Transformer model and the NeuralLoc graph neural network through learned ensembling to propose synthesis routes for small molecules. In blind tests, PhD-level chemists preferred its individual reaction predictions over those from preceding models and recorded literature reactions. The implementation and weights are open-sourced for researchers developing new medicinal molecules and materials.

AIDeveloper Mark Fenner built MiniCPM News Desk, a local-first news briefing system powered by MiniCPM5-2B. The model selects key passages from official AI and technology sources, and a rule-based editorial layer preserves dates and context before assembling a daily recap. Invalid or incomplete outputs are rejected, and the full pipeline runs locally without a hosted-model fallback.

AIIQAX has surpassed one million electronic Bills of Lading (eBLs), built on the GSBN blockchain and supporting DCSA, BIMCO, and ISO standards. The company is combining AI, IoT, and digital twins to move supply chain management from visibility toward predicting risks, with its AI-powered IoT platform covering more than 200 regions and 11,000 city pairs and about 92,000 connected devices.

AIMatei Zaharia, a prominent AI researcher, said it is very cool that GEPA works on Jev. The post is a short endorsement, linking to background about a test in which GEPA optimized Jev's prompts for extracting suspected adverse drug effects from medical sentences.

AIQwen introduced RecreationBench, a benchmark of 250 application-recreation tasks across Ubuntu, macOS, Windows, Android, and Web. Unlike GUI-only or terminal-only benchmarks, agents must explore a running reference app, recreate it in code, and pass programmatic tests plus VLM-based visual evaluation. The dataset is available on ModelScope.

AISebastian Raschka argues that Jev is more than just a classifier, since it generalizes well where earlier encoder-style classification models were usually special-purpose and limited. He suggests the main advantage lies in its data rather than the training algorithm, along with a well-designed API.

AIMuon appears to outperform AdamW because it suppresses memorization, according to WeightWatcher experiments on a single-head nanoGPT model across five seeds. At 10,000 steps, teacher-forced recall of planted sequences was about 62% for AdamW versus under 1% for Muon. The author notes that some Muon layers also show α < 2, so α alone does not explain memorization and individual layers and their ESDs should be examined.

AIGoogle Research has introduced MilleMiglia, a standardized benchmark for optimizing middle-mile logistics, the segment that moves goods across hundreds of miles overnight. The benchmark uses spatial clustering and gravity models to simulate realistic middle-mile delivery scenarios. It addresses the difficulty of optimizing these networks without public data.

AILlamaIndex launches a Parsed by LlamaParse series, using the EIA's September 2026 Short-Term Energy Outlook to show how table parsing errors can corrupt downstream data. The example value 1,186 in Table 7a means electricity sales to ultimate customers in Q3 2026, in billion kilowatthours, and misparsing its quarter, metric, or unit could flow into dashboards and forecasts. The post says LlamaParse preserves structure and footnote context needed for databases, forecasting, and AI applications.

AISemiAnalysis tested offloading DeepSeek-V4.1-Flash's Engram embedding table from HBM to host DRAM and to local SSD. On B200 configurations, DRAM delivered more total tokens per dollar and higher P90 interactivity than SSD at every measured point. The report concludes SSD offloading is likely not worth the tradeoff for production serving in its unoptimized setup.

AIGoogle's AI & Economy Research Program has added Nobel laureate Philippe Aghion as an Academic Advisor and Ajay Agrawal as a Visiting Fellow, with Anu Madgavkar and Daniel Rock named Directors. The program covers the future of work, productivity and growth, global technology diffusion, and AI's impact on scientific discovery, and builds on the recently launched AI & Economy ATLAS v1.0.

AIHamel Husain and Shreya Shankar's FAQ explains AI evals as tests of whether an AI system does what users and the business want. It recommends starting with error analysis on at least 30 traces, then turning recurring failures into binary code-based checks or LLM judges validated against human labels.

AIQianjue Technology founder Gao Haichuan argues that robotics will not see one breakthrough that suddenly lifts the whole industry, and he judges the company by deployment results rather than research papers. Qianjue, founded in 2023, has completed a Series A+ round worth a nine-figure RMB sum, with first orders coming from restaurant, cleaning, and hotel service robots. Gao says customers care about task completion, failure rates, and price rather than whether a predictive world model is used.

AIxAI released Grok Voice Transcribe 2.0, a speech-to-text model that is twice as accurate as Grok Voice Transcribe 1.0 at the same price, and ranks first for accuracy among 32 streaming models on the Artificial Analysis leaderboard. Batch transcription costs $0.10 per hour of audio and streaming $0.20 per hour, with diarization, timestamps, and key terms included. Existing Speech-to-Text API integrations gain the improvement with no code changes, and developers must pin grok-voice-transcribe-1.0 to stay on the older model during the transition.

AIGoogle Research is applying generative UI with learning design guardrails so teachers can create guided, interactive simulations for any topic and student. The post links to a sample library of more than 30 STEM interactives.

AIThe United Nations system launched UN System Data Commons, an open-source platform built on Data Commons by Google that integrates siloed global statistics into one AI-ready knowledge graph. Users can query it in natural language, browse by location or theme, and use MCP-enabled AI agents to fetch verified figures and draft charts or reports. The UN plans to add more datasets, aiming to include 80% of UN system statistical datasets by 2027.

AILlamaIndex CEO Jerry Liu spoke at Connected Stack's Founder Flash Talks about messy, complex documents that general-purpose models struggle to read, a problem enterprise agents eventually face. The company says it is building document infrastructure for agents, which it describes as the new knowledge workers.

AIDatabricks CEO Ali Ghodsi discussed leadership in an interview with @bhalligan, and the main post calls it a fun and very different conversation. The quoted background notes that Ghodsi never wanted to be CEO and was handed the interim title in 2015, when revenue was $1.5M. The source's listed takeaways emphasize focusing on the biggest bottleneck, embracing conflict, and studying competitors' weaknesses.

AIGoogle DeepMind has made its AlphaGenome Atlas freely accessible to researchers worldwide. The database is intended to help them decode the genetic causes of disease.

AIGoogle DeepMind, working with Science Stowers, built Atlas, which maps over 2,500 regulatory patterns in DNA. These sequences act as biological switches and volume dials that turn gene activity up or down across hundreds of cell types.

AIResearchers at the Broad Institute used Atlas to identify a critical mutation in the DNM1 gene, behind unexplained rare disease cases. The mutation was confirmed and validated in the lab.

AIResearchers at the University of Exeter, working with Google DeepMind, used AlphaGenome Atlas on data from more than 54,000 UK Biobank participants. The approach raised detection of rare genetic signals by over 22% and uncovered new DNA variants that influence levels of PLA2G7, a protein linked to metabolic health.

AIResearchers at the Broad Institute, the University of Exeter, and other institutions are already using AlphaGenome Atlas to identify potential disease-causing DNA variants and interpret their role. The post is a thread announcement, and it gives no further details on methods or results.

AIDeveloper Alican Kiraz fine-tuned OpenBMB's open-source VoxCPM2 voice model on nearly 100 hours of natural Turkish speech, creating Kahya-TTS for Turkish text-to-speech. The project shows how open-source voice models can be adapted to new languages and specialized datasets. The model is available on Hugging Face.

AIJetBrains describes building Air Context, a RAG pipeline that gives LLM agents semantic code search over real repositories instead of grep. The first installment covers parsing, chunking, and vectorization, arguing that fixed-size line chunks split related code and that structure-aware chunking using language grammar produces better retrieval units.

AINortheastern University MS student Soham Padia used Ai2's open Olmo 3-32B model to build Steering Arena, a public game in which players submit text prefixes to steer prosocial behavior. About 600 submissions from a few dozen people showed the top 36 entries were unreadable token strings, while the best plain-English entry ranked 37th at about 2.7 times lower score. The results suggest that once an evaluation metric is exposed, it becomes an optimization target.

AIA post estimates the cost of an RL hero run on a large open-source model at $65K per step, with a batch size of 25K. Each step takes about 2 hours, averaging 4K input and 80K output tokens.

AILlamaIndex Solutions Architect Abrar Mahi will host a webinar on turning insurance documents such as accord forms and policy documents into structured data for underwriting, policy review, and claims. The session covers extracting policy, property, and claims history into a defined schema, verifying values with citations and bounding boxes, and using confidence scores with validation rules to route items to human review.