#Data/Training

AIMoonshot AI has open-sourced MoonEP, a high-performance communication library for distributed MoE workloads. The library is designed to reduce communication overhead in expert-parallel training and inference at scale. The code is available on GitHub under MoonshotAI/MoonEP.

AIEvoCode-Bench is a multi-turn coding benchmark with 26 tasks spanning 227 sequential rounds, where agents keep a persistent workspace and must pass cumulative tests after each evolving instruction. The results show that agents perform much worse when building on their own prior work than when starting from a clean, human-completed codebase. Regressions, not failure to implement new features, are the main bottleneck, and agents that maintained a persistent requirements document more than doubled their success rates.

AIBerkeley AI Research introduces ABBEL, a framework that replaces full interaction histories with natural-language belief states that models update as new observations arrive. On CollabBench collaborative coding, belief grading closes about half the performance gap to full-context models while using fewer peak tokens and training in 50 steps instead of 100.

AIFireworks AI has made Kimi K3 available for Multi-LoRA serving and training in private preview through Fireworks Serverless Training. The post explains how small LoRA adapters can be trained on K3 and served with live merge or multi-LoRA deployment, and it reports two example tasks, Countdown and Frozen Lake, with reward curves.

AIAli Ghodsi argues that AI agents which take longer to work through a task are often worse, while Genie reaches results faster. He adds that ontology will be key to giving agents the context they need to answer correctly and quickly. The related post reports that Genie Code outperformed three general-purpose coding agents on more than 400 real user data tasks.

AIThe essay argues that Western companies increasingly rely on Chinese open-weight models like Qwen and Kimi for post-training, while Western labs cannot lawfully distill from American frontier models. It says Qwen's share of new open-model fine-tunes rose from 1% in January 2024 to 69% by February 2026, citing ATOM's Report. The authors propose controlled teacher access and tighter enforcement against foreign distillation as a domestic alternative.

AIBerkeley's STAR Lab packaged multiple LLM-based "autoresearch" algorithms into a single API within the GEPA package, letting users mix and match them. The optimizers can be applied to tasks including prompt writing, agent design, and code optimization. The quoted thread adds that GEPA, AutoResearch, and Meta-Harness each win on different tasks, and that the new optimize_anything omni meta-optimizer beats every standalone optimizer at a matched budget.

AIAl-Dahle argues that distillation is a standard training method used inside labs, under licenses, or without authorization, so it does not by itself show theft. He says a few million conversations are small against trillion-token runs, yet can matter in late-stage training, reinforcement learning bootstrapping, or training a grader. He also argues that model outputs are hard to trace after paraphrasing or mixing, and that transferred capability is difficult to measure.

AILeandro von Werra introduces The Stack v3, a dataset of 5T tokens ready for training and 120TB of raw data. He says the dataset is meant to support open code models for cyber defence, and links the download on Hugging Face.

AISequoia Capital is leading Etched's $300 million Series C at a $10 billion pre-money valuation, backing the startup's custom inference silicon. Etched says its first-generation chip taped out at TSMC on leading-edge nodes and that its first cluster ran a range of frontier models in 40 days, with early customers now getting access.

AISYNAPS-I, a multi-lab Genesis Mission project led by Lawrence Berkeley National Laboratory, uses Meta's open-source SAM 3 and DINOv3 models to segment X-ray and micro-CT scientific imagery. The fine-tuned pipeline, run on 300 A100 GPUs, reduced a grapevine xylem analysis from a month of expert annotation per time step to about 15 minutes. The team can deploy the open models inside secure national lab infrastructure, where research data must remain.

AIStanford SVL and NVIDIA Robotics introduced RoboTTT, which uses test-time training to give robot policies up to 8,000 timesteps of context at constant inference cost. The source reports that 8K-context pretraining beats 1K by 62%, and that performance keeps improving from 128 to 8K timesteps with no sign of saturation. The authors also describe one-shot imitation from human video and in-episode error recovery.

AILiquid AI released Antidoom, an open-source method that retrains a single overtrained token to eliminate "doom loops" in small reasoning models. On LFM2.5-2.6B and Qwen3.5-4B, loop rates fell from 10.2% to 1.4% and from 22.9% to 1%, respectively. The company also released IFStruct, an open-source benchmark measuring whether model outputs satisfy a schema, where LFM2.5-350M rose from 21.10% to 44.90% after training.

AIJim Fan introduced RoboTTT, a robot model that uses test-time training to compress history into a tiny inner model updated at each sensor reading. The post reports closed-loop performance rising steadily from 128 to 8K timesteps, and 8K-context pretraining beating 1K by 62%. It also claims one-shot in-context learning from human video and mid-episode error recovery, with learning continuing after deployment.

AIByteDance's UniVR-34B-Planning, built on Emu3.5 at 34B parameters, learns visual reasoning, physical dynamics, and long-term planning from visual demonstrations using a next-token objective and two-stage training on the VR-X dataset with VR-GRPO reinforcement learning. On the VR-X benchmark it scores 58.2 overall, up 18.4 points from the Emu3.5 34B baseline of 39.8. The Planning checkpoint is available on Hugging Face under CC BY 4.0, alongside a General checkpoint.

AIThinking Machines Lab says AI should extend human will and judgment, with people shaping its goals through continuous feedback rather than relying on models trained once and frozen. The company outlines three technical directions: training strong models, building tools for customization including training model weights, and developing interfaces that let personal judgment influence AI work. It also says it will publish research for the scientific community.

AIBenedict Evans argues that token prices are unstable amid a supply crunch, and that foundation models may end up as low-margin commodity infrastructure rather than holding lasting pricing power. He cites inference gross margins of 40-50% that exclude training costs, which currently exceed revenue, and compares the outlook with mobile data and semiconductor manufacturing. He concludes that the outcome remains uncertain and that value capture above the model layer would require changes not yet visible.

AICognition launched SWE-1.7, which it says reaches frontier-level coding performance at lower cost, trained from a Kimi K2.7 base. The post describes RL methods including top-p sampling replay to preserve entropy, compressed weight deltas across multi-cluster training, and self-compaction for rollouts up to six hours. SWE-1.7 is available in Devin via Cerebras at 1000 TPS.

Why it matters: The post details entropy preservation, multi-cluster weight sync, and self-compaction, offering concrete RL training techniques for long-horizon coding agents to compare against one's own pipeline.

AIBerkeley AI Research authors argue that near-free inference will make agents the dominant workload for data systems, requiring redesign for agentic speculation, agent-run state and coordination, and agent-synthesized systems. The post cites inference prices falling 9x to 900x per year with a median near 50x, and reports that about 80-90% of sub-queries in a text-to-SQL benchmark were duplicates. It frames the three directions as data systems for, of, and by agents.

Why it matters: The piece maps three concrete data-system challenges posed by near-free inference, useful for anyone designing infrastructure for agent workloads and memory.

AIMistral CEO Arthur Mensch argues enterprises should use open-source models and store their own data to avoid dependence on closed providers that retain customer data. He says companies should build continuous training loops from employee and user interactions, and shrink models to cut deployment costs. Mistral positions its Studio control plane and Forge training platform, deployed on customer infrastructure or via zero-data-retention hosting, as tools for this shift.

AIJim Fan announces ASPIRE, a system where coding agents use multimodal sensory traces from simulation and real robots to run evolutionary search over control programs and add the results to a growing skills library. The post claims up to a roughly 10x reduction in transfer learning tokens for sim-to-real and single-arm to bimanual transfer, and says the full stack will be open-sourced.

AIJohn Schulman argues that fine-tuning with the right data, such as expert judgments, can substantially outperform prompting-only approaches even as general-purpose models improve. He cites Bridgewater's work, where an expert-labeled dataset and on-policy distillation were used to fine-tune a model to triage financial documents reliably and cheaply.

AIJim Fan introduces ASPIRE, a system in which coding agents observe multimodal sensory traces and run evolutionary search over control programs to distill skills into a growing library. The post says ASPIRE shares know-how rather than pixels or weights across the sim-to-real gap, reducing transfer learning tokens by up to about 10x. The author also says the full stack will be open-sourced and provides a gallery of 150+ tasks and 90+ skills.

AIHamel Husain argues that an AI product whose output is hard to verify is a product design problem, not just an evaluation problem. He shows before-and-after sketches for an AI data agent, a PE lesson planner, and a workers' compensation report tool, each adding provenance, scoped edits, and checkable evidence. He notes that designing for verification also makes evals easier to build and grade.

AIMeta released Brain2Qwerty v2, an end-to-end deep learning pipeline that decodes sentences in real time from non-invasive brain recordings. The model reached 61% word accuracy across participants, compared with 8% for other non-invasive methods, and 78% for the best participant. Meta also released the v1 and v2 training code, and partner BCBL released the v1 dataset.

Why it matters: The source reports word accuracy and data-scaling results for non-invasive decoding, offering a benchmark against surgical brain-computer interfaces and prior non-invasive methods.

AIUnlimited-OCR, Baidu's long-context OCR model, now runs in vLLM, with a recipe provided for developers to try it. The background post says it parses entire books in one pass using Reference Sliding Window Attention (R-SWA), which keeps the KV cache fixed during decoding, and claims 35% faster throughput than DeepSeek-OCR at 6K output tokens.

AIPaddleFormers 1.2 is released with support for training DeepSeek-V4 and 128K+ long-context training. The update adds Context Parallel, Packing, Document Mask Attention, and the Muon optimizer, plus ultra-fused mHC, CSA, and HCA operators, DeepEP/HybridEP communication, and lossless FP8 training with AutoSubbatch memory balancing. The project is presented as fully open-source and is available on GitHub.

AIOriol Vinyals notes that Wojciech Zaremba wanted to remove a paper from existence about a year after publication, but says he has forgotten the reason. The post asks Zaremba to explain, and a reply describes the paper as the RNN Regularization work that improved LSTM perplexity on Penn Treebank.

AIQwen released Qwen3-ForcedAligner-0.6B-hf, a Transformers-format forced aligner that predicts timestamps for arbitrary units within up to 5 minutes of speech in 11 languages. The model accepts transcripts from any ASR system, and the documentation shows it paired with Qwen3-ASR-0.6B and NVIDIA Parakeet CTC. Until it ships in an official Transformers release, users must install Transformers from source.

AILilian Weng published a long blog post on scaling laws, which help estimate the best split of compute between data and model size before a large training run. The post covers what scaling laws predict, how compute-optimal allocation works, and why Kaplan et al. and Chinchilla reach different conclusions. It also addresses how data limits and fitting details make extrapolation difficult.

AIPP-OCRv6's recognition module uses a CTC plus NRTR dual-head design so text is decoded from visual features rather than language priors, reducing hallucination. In hallucination tests, PP-OCRv6_medium reaches 93.2%, versus 85.0% for the best VLM, and recognition accuracy across 15 scenarios is 83.2%, above PP-OCRv5_server's 78.1%. NRTR is used only during training, adding language regularization at no inference cost, and it contributes +1.16% accuracy.

AIPaddlePaddle says its PP-OCRv6_medium text detector reached an 86.2% detection Hmean in benchmarks, versus 46.8% for Gemini-3.1-Pro and 38.3% for GPT-5.5. The detector's design uses RepLKFPN with 7×7 kernels to cut FPN neck parameters from 172K to 118K, auxiliary deep supervision heads on P2–P4, and Focal Loss paired with Dice Loss, which adds +1.15% Hmean in ablation.

AILil'Log examines early empirical work showing that deep learning generalization error follows power-law curves as training data and model size grow. Hestness et al. (2017) found the exponent reflects the problem domain rather than the architecture, while Rosenfeld et al. (2020) modeled loss jointly as a function of model size N and data size D, fitting parametric forms on small configurations to extrapolate to larger ones.

AIThe post argues that China's commercial-scale EUV machines should be forecast for the late 2030s and immersion DUV for the mid-2030s, using ASML's development timeline as a reference. It also weighs factors that could push these estimates earlier or later, including state funding, espionage, talent flows, and the use of AI in R&D. The authors note that forecasts placing either milestone in the 2020s would need strong justification.

AIOpenAI reports that reinforcement learning on realistic conversations targeting traits such as honesty, epistemic humility, and corrigibility improved a model across 44 out-of-distribution alignment evaluations. Gains included reward hacking, deception, and health benchmarks, and training only on health conversations still improved non-health alignment scores. The trained model was also harder to steer toward harmful behavior with adversarial persona prompts or harmful fine-tuning.

Why it matters: The post tests whether reinforcement learning on beneficial traits in one domain transfers to unrelated alignment benchmarks and holds up under adversarial steering.

AIJohn Schulman says PPO gained a second wave in the LLM era for reasons not anticipated in the original paper. He points to the importance-ratio objective, which corrects biases from numeric error, asynchronous training, and forward-pass noise, and to the clipping objective, whose effect on entropy was unknown at publication, citing DAPO's arXiv paper.

AIOpenAI's alignment team found that re-generating 100,000 WildChat conversations with five recent OpenAI models predicted production failure rates across four orders of magnitude, with 95% of predictions within 1.04 orders of magnitude. The approach was weaker for agentic misalignment categories, where errors were about 37 times larger, and it still held roughly without access to chain-of-thought reasoning, with mean multiplicative error rising from 3.6x to 4.0x.

Why it matters: The post tests whether public chat data can predict real production failure rates, and where that prediction breaks down for agentic behavior.

AIMistral is working with companies and governments to keep their AI systems running outside external control and improving with each model release. Its Forge product enables continuous training of models based on recorded human-AI interactions, which the post calls a key unlock for efficiency.

AIMistral says it has moved its Studio deployment and Forge training products onto infrastructure it controls, decoupled from US service providers. Customers can run them in their own VPC or datacenter, or on Mistral's hosted capacity, which the company says is online and growing fast.

AIByteDance open-sourced inference code and model weights for Bernini-Diffusers, a full video generation and editing pipeline with an MLLM-based semantic planner and a DiT-based renderer. The release bundles a Qwen2.5-VL planner and Wan2.2 diffusion components in one self-contained directory, and the source recommends it over the renderer-only Bernini-R for complex instruction following.