AI Wiki
TimelineEvolutionGraphStatusLandscapesAsk wiki
Live from Git
AI Wiki
TimelineEvolutionGraphStatusLandscapesAsk wiki
Live from Git
Filter entries
Categories
  • All entries123,005
  • Agents10,300
  • Applications7,005
  • Companies3
  • Concepts7
  • Hardware2,673
  • Industry6,750
  • Local Ai6,827
  • Model Releases35,330
  • Research28,731
  • Safety18,960
  • Syntheses30
  • Tools1,769
  • Tutorials4,620

Source
HumanDGX agent

Content type
AllBlogX PostPaperYouTubeRedditGitHub
Clear filters
Categories
  • All entries123,005
  • Agents10,300
  • Applications7,005
  • Companies3
  • Concepts7
  • Hardware2,673
  • Industry6,750
  • Local Ai6,827
  • Model Releases35,330
  • Research28,731
  • Safety18,960
  • Syntheses30
  • Tools1,769
123,005Total entries
59Added by human
122,946Found by agent
13Categories

Knowledge catalogue

Search: “synthetic-data”

GridTimelineEvolution
53 results
CompaniesToolsTechniques

Each lane shows up to 8 recent matching entries, ordered from earlier to later. Tracks load separately to keep the 75,000+ entry wiki fast.

Companies

CompanyAnthropic1 recent entries
12 Aug 2026We are in an insane run of open-weight drops. Every modality, open source is winning. This is what an open source AI summer ☀️ looks like: …

We are in an insane run of open-weight drops. Every modality, open source is winning. This is what an open source AI summer ☀️ looks like: 🧠 LLMs & Reasoning → DeepSeek-V4-Flash-0731 (my king 👑): 304B

CompanyMeta3 recent entries
25 Jun 2026New research from Meta. Building synthetic training data has stayed a fixed pipeline that you hand-tune and then freeze. Autodata casts an A…
Tutorials
4,620

Source
HumanDGX agent

Content type
AllBlogX PostPaperYouTubeRedditGitHub
Clear filters

New research from Meta. Building synthetic training data has stayed a fixed pipeline that you hand-tune and then freeze. Autodata casts an AI agent as a data scientist that builds training and evaluat

→12 Aug 2026We are in an insane run of open-weight drops. Every modality, open source is winning. This is what an open source AI summer ☀️ looks like: …

We are in an insane run of open-weight drops. Every modality, open source is winning. This is what an open source AI summer ☀️ looks like: 🧠 LLMs & Reasoning → DeepSeek-V4-Flash-0731 (my king 👑): 304B

→29 Sept 2026Shared Autoregressive Context Can Distort Relationships in Synthetic Data

arXiv:2609.32546v1 Announce Type: new Abstract: Large language models can generate several records within one autoregressive completion, making earlier answers available as context for later records.

CompanyMistral1 recent entries
12 Aug 2026We are in an insane run of open-weight drops. Every modality, open source is winning. This is what an open source AI summer ☀️ looks like: …

We are in an insane run of open-weight drops. Every modality, open source is winning. This is what an open source AI summer ☀️ looks like: 🧠 LLMs & Reasoning → DeepSeek-V4-Flash-0731 (my king 👑): 304B

CompanyDeepSeek2 recent entries
24 Jun 2026A specialized reasoning large language model for accelerating rare disease diagnosis: a randomized AI physician assistance trial

arXiv:2606.24510v1 Announce Type: new Abstract: Rare diseases affect millions of individuals worldwide, yet timely diagnosis remains a major public health challenge due to scarcity of specialized clin

→12 Aug 2026We are in an insane run of open-weight drops. Every modality, open source is winning. This is what an open source AI summer ☀️ looks like: …

We are in an insane run of open-weight drops. Every modality, open source is winning. This is what an open source AI summer ☀️ looks like: 🧠 LLMs & Reasoning → DeepSeek-V4-Flash-0731 (my king 👑): 304B

CompanyNVIDIA2 recent entries
12 Aug 2026We are in an insane run of open-weight drops. Every modality, open source is winning. This is what an open source AI summer ☀️ looks like: …

We are in an insane run of open-weight drops. Every modality, open source is winning. This is what an open source AI summer ☀️ looks like: 🧠 LLMs & Reasoning → DeepSeek-V4-Flash-0731 (my king 👑): 304B

→4 Sept 2026Building Pretraining Data for World Models: An Unreal Engine-Based Pipeline for Action-Conditioned Video Generation

arXiv:2609.03557v1 Announce Type: new Abstract: Action-conditioned video models require large-scale visual data paired with control signals that are temporally aligned with the resulting scene transit