Goblin
News
AI news, distilled.
|
News
About
News
About
Filtered by:
training-and-data
Clear
Titles
Summaries
7
NVIDIA, AI4Bharat: Nemotron-pretraining-Indic-v1, a ~2.8T-Token Indic Pretraining Dataset, Releasing Soon
Open Source
1
2h ago
7
NVIDIA, AI4Bharat: Nemotron-pretraining-Indic-v1, a ~2.8T-Token Indic Pretraining Dataset, Releasing Soon
Open Source
· 1 src · 2h ago
Discuss
7
Microsoft AI: Finishes Pre-Training MAI-Base-2, 2.8T-Parameter MoE Claimed to Beat Open-Weight Models
Models
1
2h ago
7
Microsoft AI: Finishes Pre-Training MAI-Base-2, 2.8T-Parameter MoE Claimed to Beat Open-Weight Models
Models
· 1 src · 2h ago
Discuss
6
Databricks: AI Research Team Details Internal Training on AI Runtime (AIR), Reports 8x Job Growth From Agents
Products
1
2h ago
6
Databricks: AI Research Team Details Internal Training on AI Runtime (AIR), Reports 8x Job Growth From Agents
Products
· 1 src · 2h ago
Discuss
7
Meta's New Commercial Terms Allow Scraping Business-Linked Domains for AI Training
Policy
1
10h ago
7
Meta's New Commercial Terms Allow Scraping Business-Linked Domains for AI Training
Policy
· 1 src · 10h ago
Discuss
7
vLLM: Bitwise-Consistent RL Training Removes Train-Inference Drift, Holds Reward Where Baseline Collapses
Products
1
18h ago
7
vLLM: Bitwise-Consistent RL Training Removes Train-Inference Drift, Holds Reward Where Baseline Collapses
Products
· 1 src · 18h ago
Discuss
6
NVIDIA Joins European Commission's AI in Science Enablers Initiative Tied to RAISE
Infra
1
18h ago
6
NVIDIA Joins European Commission's AI in Science Enablers Initiative Tied to RAISE
Infra
· 1 src · 18h ago
Discuss
6
FactSet: Fine-Tuned Open Models Power Text-to-Formula on Databricks, Cutting Model Setup From Days to Hours
Products
1
10h ago
6
FactSet: Fine-Tuned Open Models Power Text-to-Formula on Databricks, Cutting Model Setup From Days to Hours
Products
· 1 src · 10h ago
Discuss
6
AI Runtime: Qwen 27B Fine-Tuning Cut From 7h44m to 1h17m by Fixing Kernels and Data Pipeline
Products
1
1d ago
6
AI Runtime: Qwen 27B Fine-Tuning Cut From 7h44m to 1h17m by Fixing Kernels and Data Pipeline
Products
· 1 src · 1d ago
Discuss
7
CopilotKit Releases CLiF in Early Access to Turn In-Product User Work into Agent Training Data
Products
1
1d ago
7
CopilotKit Releases CLiF in Early Access to Turn In-Product User Work into Agent Training Data
Products
· 1 src · 1d ago
Discuss
6
Apple and Partners Pretrain First Tri-Modal Masked Diffusion Models From Scratch on Text, Image, Audio
Research
1
1d ago
6
Apple and Partners Pretrain First Tri-Modal Masked Diffusion Models From Scratch on Text, Image, Audio
Research
· 1 src · 1d ago
Discuss
6
Mr.Chatterbox-2.7B-Chat: Chat Model Trained From Scratch on 19th-Century British Text
Models
1
1d ago
6
Mr.Chatterbox-2.7B-Chat: Chat Model Trained From Scratch on 19th-Century British Text
Models
· 1 src · 1d ago
Discuss
7
South Korea: Cabinet Approves AI Basic Act and Data Industry Act Amendments on National AI Data
Policy
1
3d ago
7
South Korea: Cabinet Approves AI Basic Act and Data Industry Act Amendments on National AI Data
Policy
· 1 src · 3d ago
Discuss
7
Paper: Distilling on Offline Student Rollouts (Semi-OPD) Often Beats On-Policy Distillation, With Up to 11.4x Faster Training
Research
1
2d ago
7
Paper: Distilling on Offline Student Rollouts (Semi-OPD) Often Beats On-Policy Distillation, With Up to 11.4x Faster Training
Research
· 1 src · 2d ago
Discuss
7
AI Labs Buy Workspace Data and Code From Shuttered Startups, Raising Privacy and Legal Questions
Safety
1
2d ago
7
AI Labs Buy Workspace Data and Code From Shuttered Startups, Raising Privacy and Legal Questions
Safety
· 1 src · 2d ago
Discuss
6
Apple/DTU Study: Mixing English Data Into Arabic Pre-Training Beats Hyperparameter Tuning When Data Is Scarce
Research
1
2d ago
6
Apple/DTU Study: Mixing English Data Into Arabic Pre-Training Beats Hyperparameter Tuning When Data Is Scarce
Research
· 1 src · 2d ago
Discuss
6
Flo Health: Judge-Driven Fine-Tuning on Databricks Lifts Medical Safety Score 34% to ~61% (Vendor Case Study)
Products
1
3d ago
6
Flo Health: Judge-Driven Fine-Tuning on Databricks Lifts Medical Safety Score 34% to ~61% (Vendor Case Study)
Products
· 1 src · 3d ago
Discuss
6
Federated AI Training Moves From Pilot to Production Across DOE Supercomputers, Per New Paper
Research
1
3d ago
6
Federated AI Training Moves From Pilot to Production Across DOE Supercomputers, Per New Paper
Research
· 1 src · 3d ago
Discuss
7
USA Today Co. Sues OpenAI for Over $250M, Alleging Copying of Hundreds of Thousands of Articles
Policy
1
Oct 9
7
USA Today Co. Sues OpenAI for Over $250M, Alleging Copying of Hundreds of Thousands of Articles
Policy
· 1 src · Oct 9
Discuss
6
Frankfurter Buchmesse Panel: EU AI Act Gives Publishers Tools, but Weak Enforcement Limits Value
Policy
1
6d ago
6
Frankfurter Buchmesse Panel: EU AI Act Gives Publishers Tools, but Weak Enforcement Limits Value
Policy
· 1 src · 6d ago
Discuss
7
Mecka AI Raises $60M Series B to Scale Human Motion Data for Humanoid Robot Training
Markets
1
Oct 8
7
Mecka AI Raises $60M Series B to Scale Human Motion Data for Humanoid Robot Training
Markets
· 1 src · Oct 8
Discuss
6
Databricks Blog Argues Data Access, Not Models, Is Biomedical Imaging AI's Main Bottleneck
Research
1
Oct 8
6
Databricks Blog Argues Data Access, Not Models, Is Biomedical Imaging AI's Main Bottleneck
Research
· 1 src · Oct 8
Discuss
8
Meta, Google DeepMind, and US Government Commit $1.8B to Biohub's Virtual Biology Initiative for AI Drug Discovery
Updated Oct 7
Research
5
Oct 7
8
Meta, Google DeepMind, and US Government Commit $1.8B to Biohub's Virtual Biology Initiative for AI Drug Discovery
Top
Upd Oct 7
Research
· 5 srcs · Oct 7
Discuss
7
Microsoft Research Asia Open-Sources Agent Lightning v1.0 Agentic RL Framework
Research
1
Oct 7
7
Microsoft Research Asia Open-Sources Agent Lightning v1.0 Agentic RL Framework
Research
· 1 src · Oct 7
Discuss
7
Nemotron Achieves Gold-Medal Level at Both IMO 2026 and IOI 2026
Research
1
Oct 7
7
Nemotron Achieves Gold-Medal Level at Both IMO 2026 and IOI 2026
Research
· 1 src · Oct 7
Discuss
7
SignSplit Launches with $400M Funding to Build Signed Data Infrastructure for AI
Enterprise
1
Oct 6
7
SignSplit Launches with $400M Funding to Build Signed Data Infrastructure for AI
Enterprise
· 1 src · Oct 6
Discuss
6
TII Releases Falcon-Emirati-7B: Dialect-Specialized LLM for Emirati Arabic
Models
1
Oct 6
6
TII Releases Falcon-Emirati-7B: Dialect-Specialized LLM for Emirati Arabic
Models
· 1 src · Oct 6
Discuss
6
TwelveLabs Releases Pegasus 1.6 to Convert First-Person Video into Robotics Training Data
Models
1
Oct 6
6
TwelveLabs Releases Pegasus 1.6 to Convert First-Person Video into Robotics Training Data
Models
· 1 src · Oct 6
Discuss
7
Google Launches TEE-Based Federated Learning System with Verifiable Privacy Guarantees
Research
1
Oct 5
7
Google Launches TEE-Based Federated Learning System with Verifiable Privacy Guarantees
Research
· 1 src · Oct 5
Discuss
7
Survey: 64% of Enterprises Traced Confidently Wrong AI Agent Answers to Their Own Company Data Problems
Enterprise
2
Oct 5
7
Survey: 64% of Enterprises Traced Confidently Wrong AI Agent Answers to Their Own Company Data Problems
Enterprise
· 2 srcs · Oct 5
Discuss
6
Hugging Face Launches RL Environments on the Hub to Unify Fragmented Reinforcement Learning Ecosystem
Open Source
1
Oct 5
6
Hugging Face Launches RL Environments on the Hub to Unify Fragmented Reinforcement Learning Ecosystem
Open Source
· 1 src · Oct 5
Discuss
6
UniEvo-VL: Self-Distillation Framework Lets Multimodal Models Improve Without External Teacher
Research
1
Oct 5
6
UniEvo-VL: Self-Distillation Framework Lets Multimodal Models Improve Without External Teacher
Research
· 1 src · Oct 5
Discuss
6
PewDiePie Releases Ajax AI Model for Home PCs; Claims OpenAI Banned Him for Distillation
Open Source
1
Oct 3
6
PewDiePie Releases Ajax AI Model for Home PCs; Claims OpenAI Banned Him for Distillation
Open Source
· 1 src · Oct 3
Discuss
7
Wikimedia CEO Calls Out AI Companies for Unpaid Data Use as Wikipedia Traffic Falls 8%
Policy
1
Oct 2
7
Wikimedia CEO Calls Out AI Companies for Unpaid Data Use as Wikipedia Traffic Falls 8%
Policy
· 1 src · Oct 2
Discuss
6
ServiceNow CoreAI Releases AutoSynthData for Enterprise AI Agent Training
Research
1
Oct 2
6
ServiceNow CoreAI Releases AutoSynthData for Enterprise AI Agent Training
Research
· 1 src · Oct 2
Discuss
6
Allen Institute Releases OLMo-core 3: Open Training Framework Scaling MoEs to 1 Trillion Parameters
Open Source
1
Oct 1
6
Allen Institute Releases OLMo-core 3: Open Training Framework Scaling MoEs to 1 Trillion Parameters
Open Source
· 1 src · Oct 1
Discuss
6
LIFT Transformer Architecture Propagates Hidden State Across Generation Steps
Research
1
Oct 1
6
LIFT Transformer Architecture Propagates Hidden State Across Generation Steps
Research
· 1 src · Oct 1
Discuss
6
RLTL;DR: New RL Technique Lets AI Agents Self-Improve via Self-Generated Feedback
Research
1
Oct 1
6
RLTL;DR: New RL Technique Lets AI Agents Self-Improve via Self-Generated Feedback
Research
· 1 src · Oct 1
Discuss
8
Third Circuit Upholds Thomson Reuters Win Over Ross Intelligence, Rejecting AI Training Fair Use Defense
Policy
1
Sep 30
8
Third Circuit Upholds Thomson Reuters Win Over Ross Intelligence, Rejecting AI Training Fair Use Defense
Top
Policy
· 1 src · Sep 30
Discuss
7
Google Expands Paid 'Content Offer Pilot' to Purchase Proprietary Developer Code
Enterprise
1
Sep 30
7
Google Expands Paid 'Content Offer Pilot' to Purchase Proprietary Developer Code
Enterprise
· 1 src · Sep 30
Discuss
7
Periodic Labs Deploys Periodic Neon, a 1-Trillion-Parameter AI Model, to Physical Discovery Labs
Research
1
Sep 28
7
Periodic Labs Deploys Periodic Neon, a 1-Trillion-Parameter AI Model, to Physical Discovery Labs
Research
· 1 src · Sep 28
Discuss
6
Worldmodeldata Startup Packages 1M Hours of Video Game Data to Train AI World Models
Research
1
Sep 28
6
Worldmodeldata Startup Packages 1M Hours of Video Game Data to Train AI World Models
Research
· 1 src · Sep 28
Discuss
6
FOIA Documents Reveal Federal Agencies Briefly Feared Amazon Could Train AI on Government Data
Policy
1
Sep 28
6
FOIA Documents Reveal Federal Agencies Briefly Feared Amazon Could Train AI on Government Data
Policy
· 1 src · Sep 28
Discuss
8
Oxford Quietly Let OpenAI Train on Bodleian Library's Historical Texts Without Public Disclosure
Updated Sep 28
Policy
2
Sep 28
8
Oxford Quietly Let OpenAI Train on Bodleian Library's Historical Texts Without Public Disclosure
Top
Upd Sep 28
Policy
· 2 srcs · Sep 28
Discuss
7
Trajectory.ai Introduces 'Intelligence Density' Metric and Density-Aware Training for Efficient AI Models
Research
1
Sep 25
7
Trajectory.ai Introduces 'Intelligence Density' Metric and Density-Aware Training for Efficient AI Models
Research
· 1 src · Sep 25
Discuss
7
Japanese Used Bookstores Report 5x Sales Surge Amid Suspected Bulk AI Training Data Acquisitions
Policy
1
Sep 24
7
Japanese Used Bookstores Report 5x Sales Surge Amid Suspected Bulk AI Training Data Acquisitions
Policy
· 1 src · Sep 24
Discuss
7
Research Proposes Universal Post-Training Recipe for Robotics, Introducing EXPO-FT for Stable RL
Research
1
Sep 24
7
Research Proposes Universal Post-Training Recipe for Robotics, Introducing EXPO-FT for Stable RL
Research
· 1 src · Sep 24
Discuss
7
Looped Transformer Architecture Lets 7.4B Model Match GPT-3 13B at 20x Less Compute
Research
1
Sep 24
7
Looped Transformer Architecture Lets 7.4B Model Match GPT-3 13B at 20x Less Compute
Research
· 1 src · Sep 24
Discuss
6
Meta Allows Ray-Ban AI Glasses Users to Opt Out of Visual Data Used for AI Training
Policy
1
Sep 24
6
Meta Allows Ray-Ban AI Glasses Users to Opt Out of Visual Data Used for AI Training
Policy
· 1 src · Sep 24
Discuss
6
LangChain Launches LangSmith Fine-Tuning and smithtune CLI to Build Custom Agent Models
Products
1
Sep 24
6
LangChain Launches LangSmith Fine-Tuning and smithtune CLI to Build Custom Agent Models
Products
· 1 src · Sep 24
Discuss
7
Micro1 Raises $100M+ at $4B Valuation—an 8x Jump—as AI Training Data Demand Surges
Markets
1
Sep 23
7
Micro1 Raises $100M+ at $4B Valuation—an 8x Jump—as AI Training Data Demand Surges
Markets
· 1 src · Sep 23
Discuss
Filters
Signal
Title
Category
Sources
Posted
Discuss
