Data products
Data products
Frontier Alignment
CoT reasoning, SME RLHF, SFT & red teaming
Agentic AI
Agent trajectories, RL environments & tool-use logs
Speech & Audio
Expressive TTS, emotion detection & dialectal speech
Multimodal AI
VLM data, video annotation & cross-modal alignment
Physical AI
LiDAR annotation, robotics trajectories & sensor fusion
Model Integrity
Hallucination benchmarking, bias & compliance audits
Why Appen
30 years of AI data expertise
SOC 2 & ISO 27001 certified
1M+ vetted contributors
500+ global locales covered
Independent model evaluation
Perspectives
Content
Blog
Expert commentary on frontier AI development
Research
Research papers and publications from Appen
Case Studies
How leading AI labs work with Appen
Webinars & Events
Live and on-demand sessions with our experts
Featured
Why Human Evaluation Still Wins
Master RLVR at Scale
Improving LLM Safety Evaluations
About
Company
Annotation Platform
AI Data Platform (ADAP) — automation meets human oversight
About Us
30 years pioneering AI data
Investor Relations
ASX-listed: APX
News
Latest announcements and press
Careers
Build the future of AI with us
Get in touch
Join our workforce
Join our workforce
Get in touch
Home
Resources
Press External
A startup claims it broke through a bottleneck that’s holding back LLMs
Press
A startup claims it broke through a bottleneck that’s holding back LLMs
Appen
•
Jun 2026
Read article
MIT Technology Review
