Corrections
When we get something substantively wrong, we fix it and log the fix here; small copyedits live in each article's revision history instead. Found an error? Use the "Report issue" button on the article (no account needed). So far we have received 3 reports and reviewed 1, taking a median of 4.8 days. 2 are waiting, the oldest since September 26, 2026. That is our record rather than a promise. See the verifiability policy for how checking works.
321 articles were corrected in the last 30 days. Every correction note and resolved report is listed below, newest first. A fix applied to several articles with the same note is shown once, listing each article; "See the change" opens that version in the article's revision history.
- editorial
Consolidate duplicate NVIDIA HGX coverage; restore the family's changing physical forms, document topology, integration, certification, fabric operations, and partitioning, and qualify Rubin status and conflicting B300 memory figures.
- editorial
Retitle and refocus the duplicate Groq page as the hardware, architecture, interconnect, and compiler-stack article; replace stale company categories with exact hardware, inference, and infrastructure taxonomy; separate first-generation and Groq 3 specifications; add reciprocal company and processor disambiguation; and replace promotional comparisons with primary and independent evidence.
- editorial
Separated the Groq company page from Groq hardware and replaced unsupported or stale claims with verified corporate history, financing, licensing, leadership, cloud operations, and partnerships.
- editorial
Restores the AI chip title and separates physical semiconductor devices, packaging, specifications, and supply from the broader AI accelerator system category.
- editorial
Separated AI accelerator from AI chip and replaced overlapping taxonomy and volatile product claims with verified architecture, workload, software, and benchmarking guidance.
- editorial
Clarified Absolute Zero Reasoner's zero-data boundary, seeding, rewards, evaluation, release status, and limitations; added primary and independent sources.
- editorial
Expanded Online-Mind2Web's sourcing and corrected its taxonomy, dataset lineage, evaluation protocol, result boundaries, update history, and licensing through July 28, 2026.
- editorial
Clarified FormulaOne's dataset tiers, public-test boundary, original model settings, and theoretical scope; added dated results and supporting references.
- editorial
Expanded WindowsWorld with its ACL 2026 methodology, scoring, judge validation, model-specific results, comparison to later benchmarks, and release-state reproducibility limits through July 28, 2026.
- editorial
Removed legacy MediaWiki markup and a faux infobox; corrected FANUC governance, history, reportable-segment framing, current financials, robot market-share claims, customer assertions, and physical-AI shipment language through July 28, 2026.
- editorial
Separated the original 2005 corrected estimator from modern direct mixed-penalty implementations; corrected parameter conventions, grouping and p>n boundaries, coordinate scaling, solver chronology, cross-validation, standardization, and comparison claims through July 28, 2026.
- editorial
Removed the legacy faux infobox, unsupported forecasts, rumored model specifications, and misattributed research and litigation; rebuilt OpenAI's history, governance, products, financing, safety, and legal record through July 28, 2026 from primary and strong independent sources.
- editorial
Reframed Baidu AI as an umbrella rather than a standalone division; separated ERNIE, Qianfan, PaddlePaddle, Kunlunxin, and Apollo; replaced unsupported rankings and stale figures with bounded evidence through July 28, 2026.
- editorial
Removed the legacy faux infobox and External links section; separated Alibaba Cloud, Cloud Intelligence Group, Qwen, PAI, Model Studio, and ModelScope; corrected market, revenue, region, chip, and medical-study claims through July 28, 2026.
- editorial
Reverified SlidesGPT's legacy custom GPT, current ChatGPT app, operator, chronology, access, features, pricing, data retention, terms, and API status through July 28, 2026, while removing the copied system prompt and unsupported availability, model, and comparison claims.
- editorial
Removed illegal frontmatter and H1; corrected Adafactor's memory accounting, momentum-state formula, T5 and PaLM schedules, and implementation differences; bounded adoption and convergence claims through July 28, 2026.
- editorial
Reverified Neurealm's corporate lineage, May 2025 rebrand, acquisitions, leadership, NeuGAIN platform, functional-safety claims, and NVIDIA Halos role through July 28, 2026, while distinguishing company statements, inspection accreditation, and final certification.
- editorial
Reverified Microsoft Agent Framework's release chronology, architecture, language and package scope, predecessor-project status, interoperability, licensing, support, and security limitations through July 28, 2026.
- editorial
Expanded Deepu Talla's verified biography through July 28, 2026; clarified his education, Texas Instruments and NVIDIA role progression, robotics portfolio, safety work, speaking, recognition, and philanthropy while removing a promotional quotation and bounding company claims.
- editorial
Reverified Kimi K3 through July 28, 2026; corrected its release state, active parameter count, architecture, modalities, API behavior, pricing, license, benchmark framing, and independent safety evidence while removing unstable finance, market, service-rank, and launch-hour claims.
- editorial
Reverified Kimi's history, product scope, model lineage, licensing, and safety evidence through July 28, 2026; removed unsupported funding, usage, revenue, pricing, and benchmark claims.
- editorial
Rebuilt the article from immutable version 13 with 44 resolved footnotes; removed all legacy MediaWiki import markup; corrected the pending 2026 financing and IPO reports, open-weight licensing, K3 release status, and unsupported model-size and training-cost claims; and preserved every existing internal-link target.
- editorial
Rebuilt the references as 54 resolved footnotes; corrected AlphaGenome's evaluation count, the tokamak date, the Frontier Safety Framework version, and July 2026 Gemini and robotics lineups; removed unsupported headcount, office, ranking, acquisition-price, and forecast claims; and preserved the verified history, research scope, and every existing internal-link target.
- editorial
Connected the source to inline claims and expanded the article with verified coverage of media production, distribution, synthetic content, labor, rights, and governance
- editorial
Added a complete source-bound references section; repaired malformed, duplicated, ambiguous, and incorrect AI acronym expansions; added current high-use abbreviations; reorganized the glossary for readability; and preserved verified definitions and useful internal links.