#Data/Training

AIApple researchers introduce RISED, a framework that uses rubrics to guide data selection and policy supervision when training one LLM agent across multiple interactive environments. An LLM judge tags rollouts with a shared rubric vocabulary, positive rubrics provide privileged context for an on-policy self-distillation teacher, and negative rubrics steer generation away from recurring failures. The authors report that RISED achieves the highest mean pass rate across environments and ranks first or second in each environment, across model backbones.

AIDust is a zeroth-order method that pretrains transformers and sometimes matches or exceeds backprop given large compute. The authors report it is about 1,000 to 10,000x more compute efficient than EGGROLL, the state-of-the-art ES method, for training transformers. The post also cites the gradient-alignment result up to 1B tokens and the virtual population idea for scaling.

AIPyTorch has consolidated all media decoding and encoding for images, video, and audio into TorchCodec, which now runs on CPU and CUDA. TorchVision and TorchAudio are narrowed to focus on their transforms, with models, datasets, and pipelines no longer under active development. All three libraries are now ABI stable and no longer need rebuilding for each PyTorch release.

AIHarrison Chase praises Cognition's "Dreaming" feature, which lets Devin clean stale memory records and surface latent information offline. He argues agent memory needs an offline cleanup loop rather than only better retrieval, and questions how inferred memories get validated before use. He also welcomes Cognition's plan to release Agent Memory Repo as an open standard.

AIReflection AI says its Beam model was pretrained in 4 weeks on 24T high-quality tokens, giving it innate coding capabilities. The company credits MoE stability improvements and large-scale data curation and deduplication for a base model it claims outperforms open-source base models of the same class. It presents this strong reasoning foundation as what makes sustained reinforcement learning gains possible.

AIReflection says it ran Beam, its reinforcement learning system, on 10.5k GB300 GPUs for four weeks, which it describes as the largest publicly documented RL run it knows of. The company credits algorithmic advances combined with distributed infrastructure for making the system scale. Across its eval suite, capabilities kept improving as RL increased, with no sign of a plateau.

AILlamaIndex argues that traditional OCR, which makes one pass and returns unchecked text, is being replaced by agentic OCR. The approach treats parsing as a loop with layout-aware reading order, routing of hard elements such as tables and charts to suitable models, and multi-pass self-correction.

AILiving Models paired Google's Gemma 4 with BOTANIC-1, a plant-DNA model trained on 320 species, to identify causal genetic variants. In a melon yield test, the pipeline ranked the target mutation first out of 2,494 possibilities in under four minutes. The approach aims to speed up breeding of climate-resilient crops that would otherwise take years of field trials.

AIEra, launched by Ofir Ehrlich's team, generates a complete simulated company spanning Salesforce, Slack, Jira, Zendesk, Gong, and Deel, plus cloud databases and storage. Agents interact with it through live MCP and API interfaces, and because Era generated the company, it knows the exact ground truth for testing and benchmarking. The post says the product is live today and free.

AIA survey of 300 data, AI, and technology executives found only 34% of organizations' agentic AI projects reach production, with legacy systems, security concerns, and missing knowledge context as main obstacles. Production leaders, who advance 61% of projects beyond pilot, show stronger semantic knowledge capabilities. Most firms plan to invest in retrieval pipelines, AI-ready APIs, retrieval-augmented generation, and knowledge graphs.

AIOpenAI is rolling out text watermarking for EU AI Act compliance, with opt-in access for API customers globally on select models starting today. Watermarking stays off by default in the API, while an invisible watermark will be added to eligible ChatGPT and Codex text in the European Union over the coming weeks. Access to the text watermark detector is initially limited to approved researchers and expert organizations, and the image and audio verification tools remain publicly accessible.

AIHugging Face says a capture proxy lets reinforcement learning train open models inside unmodified coding harnesses such as Claude Code, Codex, and OpenCode. The proxy records the exact token IDs and logprobs vLLM samples and hands them to TRL for training. On LFM2.5-2.6B, training in four harnesses at once raised OpenCode results from 34% to 58%, while SFT on 3,189 Qwen3.8-27B rollouts plateaued at 47.5%.

AIEmmy-winning Light Sail VR is using AMD Agent Computers with private, local AI on Ryzen AI Max and Radeon AI PRO R9700 to oversee up to 9 projects simultaneously. The studio says this frees more time for storytelling.

AICloudflare announced 46 products and updates during Birthday Week 2026, including the cf CLI for the entire Cloudflare API and EmDash, an open-source Astro-based serverless CMS whose plugins run in isolated Worker sandboxes. The company also said it plans to become a public certificate authority that issues free Merkle Tree Certificates for post-quantum authentication.

AIA native word-level transcription model outputs structured, timestamped arrays of word, spacing, and audio_event tokens directly from audio input, without a secondary forced-alignment pass. Audio events such as laughter or applause are tagged separately, which the source says helps with captioning, searchable archives, and highlight identification. The source notes Scribe's word-level transcription supports up to 5 independently transcribed channels.

AISmartphone prices have risen about 15% globally this year, and newly launched models cost roughly 25% more than last year, as a memory chip shortage driven by AI data center demand raises manufacturing costs. Shipments of sub-$100 smartphones fell almost 60% year over year in the second quarter of 2026, according to IDC, and Chinese makers are cutting entry-level projects in favor of pricier devices. GSMA warns the trend could widen the digital divide.

AIAI image models default to attractive women because training data, averaged-face aesthetics, and user preference feedback reinforce one another. A 1973 test image from Playboy, later widely used in image processing, shows how such defaults form early. Reward models trained on user choices can increase NSFW output even when prompts are unrelated.

AIPromptArmor reports that a malicious Skill can make Databricks Genie Code display a phishing modal and exfiltrate tenant data without human approval. The attack exploits Skills loaded from users' personal workspaces and a display interface that lacks egress controls, and Databricks, after disclosure on August 16, 2026, said users are responsible for ensuring uploaded Skills contain no malicious content.

AIApple and Stanford researchers built two open-ended probes using a Wizard of Oz technique so participants could train personalized machine learning systems on phenomena they defined themselves. In a week-long exploratory study, participants identified four sites where ontological boundaries were negotiated: the boundaries of a phenomenon, the subject as part of relations, signal versus noise, and the objectivity of data. The paper offers starting points for supporting boundary negotiation through design.

AIEmpero announces Homebrew, an open agent that builds AI models for users. According to the post, users describe what they want, and the agent picks a base model, prepares data, sets parameters, and runs training. The project is available on GitHub at empero-org/homebrew-ai.

AIDemis Hassabis says he is proud of Google's work using AI to accelerate science and medicine for society's benefit. The quoted post from Sundar Pichai highlights recent examples, including the open AlphaGenome Atlas mapping 9B possible single-letter genetic changes and the WeatherNext 3 global weather model. It also points to translation services now available in nearly 300 languages.

AILlamaIndex introduced Extract v2.5, a series of agents tuned for document extraction, including cost-effective, agentic, and agentic plus tiers, available in LlamaParse. The company says the agents reduce error rates by 2x or more compared with frontier models at a small fraction of the price, and they handle handwritten and drawn annotations on scanned documents while grounding values in the source text.

AIAlibaba T-Head unveiled the Zhenwu V900 at the Apsara Conference 2026, featuring 216 GB of memory capacity and 1,200 GB/s of interconnect bandwidth. The V900 is slated to deliver 3x the performance of the Zhenwu M890, with shipments starting in Q1 2027.

AIReplit CEO Amjad Masad argues that general models could train smaller, domain-specific successors on the fly when they detect a limited use case. He compares this to a just-in-time compiler that emits optimized code during execution. He says such specialized models could be cheaper, less vulnerable to prompt injection, and less harmful than general agents.

AIIndexTeam has published an official NVFP4 (W4A4) quantized version of its Index-Echo-S2TT-2B speech-to-text translation model on Hugging Face. Only the text LLM backbone is quantized, while the audio tower, connector, and speech-synthesis components remain in BF16. Perplexity rises 5.80%, from 4.8772 to 5.1599, on a fixed corpus, and full FP4 speedup requires an NVIDIA Blackwell GPU.

AIThe post argues that aligning future superintelligent models by controlling which ideas enter pretraining is fragile and resembles witchcraft more than engineering. It concludes that this approach cannot serve as the basis for the safety of future models.

AIDatabricks has rolled out a set of updates to Genie One spanning context, data access, collaboration, and automation. Genie Ontology is enabled by default to provide business-aware context, and workspace instructions can apply organizational data conventions to every prompt. Users can also upload Word documents, images, CSVs, spreadsheets, and PDFs, query Unity Catalog tables with schema preview and one-click access requests, and automate recurring work with scheduled tasks that reference past runs.

AICohere reports that after the first training step, its model could already translate over 90% of the training documents, which created a data problem. Kocmi describes how the most difficult samples were used to strengthen North Small Translate's capabilities. This post is part 3 of a six-part thread.

AICohere says its supervised fine-tuning extended machine translation with data focused on post-editing, error detection, and terminology. The team also ran multiple reinforcement learning and DPO stages aimed at errors the model showed.

AISebastian Raschka released round six of his Reasoning from Scratch series, introducing Reinforcement Learning with Verifiable Rewards (RLVR) and Group Relative Policy Optimization (GRPO) with an implementation. The video covers accuracy and format rewards, DeepSeek-R1 training, and GRPO versus PPO, then walks through a training loop and evaluates checkpoints on MATH-500.

AILlamaIndex introduced Extract v2.5, a set of document extraction agents that can reconstruct records split across pages and assemble them with thousands of other cells into structured tabular output. The post says the agents handle real-world documents like insurance claims, regulatory filings, and legal schedules, where a record may start on one page and finish on the next. The accompanying background post claims record-spanning-page accuracy rose from 85.5% to 96.5%, and that the agentic tier outperforms Opus 5.5 and GPT-6 Sol at 30% to 4x lower cost.

AICMU researchers released SMDD-Bench, a benchmark of 502 small-molecule drug design tasks that use RDKit, ADMET-AI, and Boltz-2 as feedback loops. The authors argue that long-horizon planning, exploration, and learning from imperfect feedback remain open problems beyond math and coding, and the benchmark is available in Prime Intellect's Environments Hub for training with prime-rl.

AIReplit chat now generates interactive charts when users ask Replit Agent to visualize data. Users can also choose GPT-6.1 Sol from OpenAI or Claude Sonnet 5.5 from Anthropic when building with Agent, or stay in auto mode. Jev is available through Replit AI Integrations for classifying content, routing requests, and scoring leads without managing API keys.

AIPrime Intellect says long-context agent serving depends on retaining history, scheduling new work, and moving cached state efficiently. It optimized three paths separately: prefill topology and scheduling, compressed KV with a fused attention kernel, and a transfer-friendly cache layout.

AIThe vLLM team integrated Helion, a PyTorch-native kernel DSL, into vLLM's linear backend, using per-shape autotuning to select among Standard GEMM, Split-K, and Swap-AB variants. On NVIDIA Hopper GPUs, the Helion backend outperformed the default CUTLASS and DeepGEMM backends across the evaluated models, with more than 10% throughput gains for some workloads. The work focuses on FP8 and INT8 quantized GEMM.

AIMeta researchers used Muse Spark to prove a clear rule for when a cycle-based relaxation of a hard optimization problem matches the original exactly and when it leaves a gap. The work concerns completed length-three alpha cycles, and the full paper is linked in the post.

AIEpoch AI estimates that compute built from projected 2025 to 2027 high-bandwidth memory shipments could support tens to hundreds of millions of frontier AI agents, or billions of cheaper ones. Running nonstop, the top-tier agents would match the working hours of 140 million to 700 million full-time employees, and the central DeepSeek V4 Pro estimate of about 1.9 billion agents would match 8 billion workers.

Why it matters: The estimate converts memory shipments into agent capacity and revenue ranges, showing how hardware supply could translate into labor and sales if demand keeps up.

AIAdaption Labs has made Invent, its tool for generating AI training datasets from a plain-language description, available through an API. Developers can reportedly produce AI-ready training datasets in minutes with a few lines of code, according to the quoted post. Documentation is available at docs.adaptionlabs.ai.

AINVIDIA reports that fine-tuning Nemotron 3.5 ASR reduced word error rate on Najdi and Hijazi Saudi Arabic from 55% to 30%. The post says a new tutorial shows how to adapt the model for other dialects and languages.

AINathan Lambert and Tom Zick have unveiled Trillium Labs, a new non-profit focused on the open science of frontier AI. The lab plans to build open post-training recipes and expand into open infrastructure to study topics such as RSI, reward hacking, and multi-agent systems. It is hiring, fundraising, and seeking compute, with support from Halcyon Futures and Schmidt Sciences.