Loading…
Topic
617 posts about Machine Learning. Every summary links to the original.
See the idea
A walkthrough of how a 3D CT model turns a voxel volume into a structured diagnostic report: isotropic resampling, 3D patch tokens, spatial encoding, and language-model reasoning.
The Good Stuff
Paid picks we use ourselves. Links carry a referral code — a subscription through one earns us a commission at no extra cost to you.
Newest first
amazon.science
AI agent performance increasingly depends on the harness mediating interactions between language models and tools rather than raw model reasoning. To address mismatches between model intent and harness execution, the authors developed Simple Strands Agent, an open-source single-agent harness. The architecture implements robust tool-interface practices, such as condensing middle log outputs from bash commands, enforcing line-aware string replacements with ambiguity checks, and returning diff files after edits for verification. Evaluated across benchmarks like SWE-Bench-Verified, SWE-Bench-Pro, and Terminal-Bench 2, these model-harness alignments yielded consistent performance gains across frontier models without task-specific tuning. The findings show that model families require tailored interface codesign alongside stable evaluation infrastructure to achieve optimal benchmark reliability.
OpenEnv is becoming a community-coordinated, open-source tool for building agentic execution environments such as terminals and browsers, with the stated goal of helping train open-source agents. Starting now, a committee including Meta-PyTorch, Reflection, Unsloth, Modal, Prime Intellect, Nvidia, Mercor, Fleet AI, Microsoft, Hugging Face, and RadixArk will coordinate the project, which now lives at huggingface/OpenEnv and is supported or adopted by numerous AI organizations. Technically, OpenEnv is being narrowed to an interoperability and protocol layer: environments expose a Gymnasium-style reset(), step(), and state() API through client/server deployments, standard HTTP or WebSocket protocols, Docker packaging, and MCP compatibility. It does not define rewards or training loops; those remain in specialized libraries, while OpenEnv connects harnesses, environments, and trainers across ecosystems. Planned work includes external rewards, dataset-backed tasksets, harness integrations, end-to-end examples, and automatic environment validation.
ClickHouse announced an expansion of its global go-to-market leadership team, appointing Ed Lenta as Vice President, Asia Pacific and Japan (APJ). Based in Singapore, Lenta will lead regional go-to-market efforts and accelerate adoption of ClickHouse Cloud and the open-source database, following leadership roles at Databricks, AWS, and VMware. The company also appointed Takeshi Kaneko as Country Manager of Japan and added executives for financial services, public sector, global solutions architecture, and revenue strategy and operations. ClickHouse said the appointments support its growth after surpassing $250 million in annual run-rate revenue, more than triple the prior year, and exceeding 4,000 customers, including more than 1,000 net new customers in one quarter.
Explainable AI (XAI) comprises techniques that help people understand how AI systems produce specific outputs, particularly when machine-learning and deep-learning models operate as black boxes. It distinguishes intrinsically interpretable models, such as decision trees and linear or logistic regression, from post-hoc methods including SHAP, LIME, counterfactuals, saliency maps and Grad-CAM. A typical workflow selects a model and prediction, applies a method suited to the model and audience, reviews outputs such as feature scores or heatmaps, and uses them to assess accuracy, fairness, reliability and compliance. The article stresses that post-hoc explanations are approximations rather than definitive proof, so teams should validate them with domain expertise and, where appropriate, combine methods; MLflow and Unity Catalog can preserve explanation artifacts, lineage and auditability.
The playbook presents AI and data transformation as an enterprise capability built on trustworthy, well-governed data, rather than a model-only project. It explains how governance, scalable pipelines, and continuous quality monitoring address incompatible formats, missing values, duplicate records, and schema inconsistencies from systems such as CRM platforms, operational databases, IoT sensors, and cloud applications. Operational guidance covers KPI design, named pipeline ownership, review checkpoints, data mapping and lineage, automated de-duplication, deterministic enrichment, regulatory classification, and role-based access. It states that deterministic tests and human review remain necessary for AI-assisted code, while telemetry and quarterly audits help detect reliability, governance, and data quality degradation. The concluding position is that operational discipline—not model sophistication—supports reliable machine learning, predictive analytics, and generative AI outcomes.
Nemotron 3.5 Content Safety is NVIDIA’s 4B-parameter model for evaluating user prompts, optional images, and optional assistant responses together, with multilingual coverage and enterprise-specific policy enforcement. It extends Nemotron 3 by accepting natural-language custom policies, producing optional auditable reasoning traces, and supporting binary verdicts, category-labelled verdicts, or THINK mode; disabling THINK preserves the default low-latency path. Built on Google Gemma 3 4B IT with a LoRA adapter and a 128K context window, it uses the Aegis 2.0 taxonomy of 13 core categories and 10 subcategories, and targets deployment on GPUs with 8GB or more of VRAM. The accompanying multimodal, multilingual dataset includes concise reasoning traces, while reported results average 96.5% on Multilingual Aegis, 88.8% on RTP-LX, and about 85% across the evaluated benchmark set.
Retailers face major operational bottlenecks when managing supply chain disruptions through legacy enterprise resource planning systems and siloed documents. Traditional tabular structures struggle to index multi-modal data like images and PDF contracts, delaying critical supplier discovery during regional crises or tariff changes. To overcome these constraints, organizations decouple supplier management into a dedicated operational data layer using MongoDB Atlas and Voyage AI. This architecture stores polymorphic supplier profiles and high-dimensional vector embeddings within a unified collection, enabling natural language semantic search for alternate partners. Real-time updates propagate across systems using MongoDB Change Streams, allowing businesses to rapidly assess supply chain impacts and identify alternative vendors in minutes.
amazon.science
Evaluating search-augmented AI research reports requires verifying complex, multi-source claims that depend heavily on surrounding context. Unassisted PhD-level domain specialists tasked with establishing static ground-truth labels achieved only 60.8% accuracy on a hidden test set during controlled testing. To resolve benchmark errors on cognitively demanding tasks, Amazon AGI introduced the audit-then-score evaluation protocol alongside DeepFact-Bench and the DeepFact-Eval verification system. Under this protocol, model disagreements serve as challenges requiring concrete counterevidence and written rationales, enabling human auditors to evaluate competing claims and update benchmark ground truth prior to scoring. Operating under this iterative workflow, expert test accuracy rose to 90.9% across four rounds, and DeepFact-Eval reached 83.4% accuracy with GPT-4.1.
DharmaOCR applies Direct Preference Optimization (DPO) beyond chat alignment to reduce text degeneration in structured OCR, where outputs are either acceptable transcriptions or failure loops. In a benchmark of Brazilian Portuguese structured-document OCR, vanilla degeneration rates among open-source model families ranged from below 1% to above 33%, while supervised fine-tuning (SFT) did not consistently reach production-acceptable levels. After SFT, the pipeline generated multiple candidates for 23,726 documents, used an automated LLM judge to score them, and retained the model’s degenerate outputs as rejected examples paired with chosen outputs. DPO reduced degeneration in every tested family, by 59.4% on average and up to 87.6%, without reducing extraction quality according to the benchmark; applicability is presented as dependent on failures being categorically identifiable, reliably scoreable, and sufficiently numerous.
Ramp’s Stack is an AI-native accounting suite designed to automate repetitive book-closing work, including reconciliations, variance analysis, data entry, and schedules and accruals. To avoid overfitting to individual design partners, Ramp built a benchmark of synthetic business worlds, realistic accounting tasks, and accountant-written grading criteria, including standard and roll-forward worlds for testing memory transfer. The benchmark contains 237 tasks and 3,469 grading criteria in the analyzed slice, and supports repeated runs to compare models, prompts, tools, skills, harness changes, and memory behavior. Optimization included ablating skills, shrinking a spreadsheet skill from 14,000 to 5,000 characters, and tuning the system against end-to-end task performance rather than narrow evaluations. The resulting Stack system achieved the highest agent performance, with 4% higher accuracy and 3% better Pass@1, while remaining on the latency frontier with GPT 5.4; schedules and accruals were harder for raw models than variance analysis.
When historical travel data became uninformative during the asynchronous post-COVID recovery, Airbnb redesigned its demand forecasting to borrow signals across geographies rather than waiting for local data. Using a hierarchical Bayesian framework, the team propagated updated posterior estimates from early-recovering corridors as informative priors for structurally similar, later-recovering corridors. This allowed accurate corridor-level demand forecasting in near real time during periods of high disruption and uneven market reopening.
Data analytics and data science are compared as data-focused career paths, with analytics interpreting existing, usually structured data for trends and business decisions, while science builds models and automated systems to predict future outcomes. The guide contrasts typical questions, outputs, tools, education paths, and roles: analysts use SQL, Excel, Tableau, or Power BI for dashboards and reports, whereas data scientists use Python, R, Apache Spark, and MLflow for predictive models and algorithms. It describes analytics types from descriptive through prescriptive, and a data science workflow spanning collection, feature engineering, training, validation, and deployment, including unstructured data such as text, images, and sensor streams. It also explains collaboration, including analysts defining problems and baselines before scientists build models, and offers portfolio suggestions and questions for choosing a path.
The article presents practical data lakehouse patterns for streaming analytics, IoT pipelines, machine learning workflows, and enterprise reporting, addressing the gap between theoretical definitions and deployable examples. It explains how a lakehouse combines low-cost, schema-flexible object storage with schema enforcement, ACID transactions, data versioning, lineage tracking, and query performance, allowing SQL and ML workloads to use shared open-format tables. Examples include second-level fraud detection, medallion-based historical analytics, predictive maintenance from sensor data, and governed customer 360 profiles. The implementation guidance covers raw storage and partitioning, centralized catalogs, decoupled compute, role-based access, time travel, migration coexistence, SLAs, observability, and lifecycle policies. Together, these patterns are presented as a unified architecture that reduces duplication and data movement while supporting governed analytics at scale.
Holo3.1 is a family of computer-use models released to improve robustness across web, desktop, and mobile environments, agent frameworks, and deployment targets. Based on Qwen, it adds native function-calling alongside structured JSON outputs, expands mobile automation, and offers 0.8B, 4B, 9B, and 35B-A3B sizes for different cost, latency, and performance needs. On AndroidWorld, the 35B-A3B model rises from 67% to 79.3%, while the 4B and 9B variants rise from 58% to 72%; function-calling and native execution achieve near-parity across the stated benchmarks, and Holo3.1 improves more than 25% over Holo3 in the Holotab harness. The release also provides 35B-A3B FP8, Q4 GGUF, and NVFP4 checkpoints for local inference: NVFP4 matches FP8 on OSWorld at about two points below BF16, reaches 1.74 times BF16 token throughput on DGX Spark, and with harness optimizations cuts average step time from 6.8 seconds to 3.3 seconds.
The guide frames enterprise choice between retrieval-augmented generation (RAG) and fine tuning as a trade-off between injecting knowledge at inference time and changing model weights before deployment. RAG connects an LLM to external stores by embedding a query, searching vector databases for similar chunks, and adding retrieved context to the prompt, making current information and citations possible. Fine tuning uses curated input-output examples and supervised training to update parameters; full-model training is expensive, while LoRA trains a smaller set of added weights and can support a single-GPU pilot. The decision depends on whether the priority is stable behavior and output format or fresh, traceable knowledge, alongside data quality, latency, compute, and maintenance requirements. For many mature deployments, the guide recommends combining both: fine tuning for domain understanding and consistency, and RAG for real-time data.
JetBrains is releasing Mellum2, a 12B-parameter Mixture-of-Experts model trained from scratch on natural language and code. It targets latency-sensitive, high-throughput workloads including routing, retrieval-augmented generation (RAG), summarization, sub-agents, and coding features. The model has 12B total parameters but activates 2.5B per token, with its architecture keeping total capacity high while making inference more efficient for real-time use. Released under the Apache 2.0 license, Mellum2 is described as competitive with similarly sized open models on code generation, reasoning, science, and math benchmarks while delivering more than 2x faster inference. Its text-and-code focus supports high-throughput production workloads and private deployments where proprietary code or internal data may be involved.
Executable UDFs are available in public beta on ClickHouse Cloud, allowing Python functions packaged as zip files to be called from SQL like built-in functions. ClickHouse runs them in a pool of long-lived sandboxed processes, keeping model code near the data and supporting use in ad-hoc queries, joins, and materialized views. The demonstration wraps a roughly 270K-parameter PyTorch autoencoder in an embed_trade UDF that consumes 14 features, produces a 32-dimensional embedding and reconstruction error, and scores trades as they enter through a materialized view. Training uses streamed Parquet chunks, incremental StandardScaler fitting, early stopping, and saved model artifacts; the deployed pool keeps the 2MB model hot, so inference occurs once per trade and downstream queries read the resulting table.
CMS’s Transforming Episode Accountability Model (TEAM), beginning January 1, 2026, requires selected organizations to manage total cost and quality across five surgical episode categories and a 30-day post-discharge window. The post argues that retrospective dashboards and delayed claims data cannot support intervention or prepare hospitals for downside risk, especially amid complex integration across EHR, claims, post-acute, social-determinants, and provider data. It proposes a modern, cloud-native lakehouse foundation with unified data, operationalized AI/ML, embedded clinical workflows, and scalable architecture. Recommended capabilities include risk stratification, next-best-action decision support, post-acute optimization, provider benchmarking, and predictive performance alerts. In a composite 500-bed medical-center example, the approach is associated with 15% lower SNF costs, 8% better risk-capture completeness, and 12% fewer readmissions.
The post introduces torch.profiler through a beginner-oriented walkthrough of profiling a matrix multiplication and addition, explaining how to configure the profiler and interpret its outputs. It uses a small function with torch.matmul and torch.add, annotates it with record_function, captures CPU and CUDA activities after warm-up iterations, and exports a profiler table plus a Chrome trace. The table summarizes event counts and CPU or GPU time, while the trace shows execution order, call hierarchy, CPU-GPU gaps, kernel launches, and overlap. Comparing 64x64 and 4096x4096 matrices demonstrates a shift from overhead-bound behavior to compute-bound behavior; the guide also covers dispatch chains, hidden device-to-device copies, profiler buffer activity, and torch.compile's runtime and caching behavior.
amazon.science
Amazon Research Awards has announced 70 award recipients across 49 universities in 11 countries for its fall 2025 funding cycle. The funded research spans six distinct tracks: AI for Information Security, Agentic AI, Automated Reasoning, AWS Cryptography, Cybersecurity and Anti-Abuse Technologies, and Sustainability. Each selected academic team receives unrestricted financial grants, AWS Promotional Credits to access specialized cloud AI and machine learning services, and access to more than 700 Amazon public datasets. Awardees are paired with an Amazon research contact who provides advisory support, consultation, and access to company training sessions. Furthermore, the program explicitly encourages recipients to publish their discoveries, deliver presentations at Amazon offices worldwide, and release associated software implementations under open-source licenses.
The best engineering reads, once a week.
