Sequoia
·
4 months ago
Edra, founded by former Palantir executives Eugen Alpeza and Yannis Karamanlakis, creates dynamic knowledge bases for AI agents by analyzing existing company data like support tickets and emails rather than requiring manual documentation. Early use cases focus on IT service management and customer technical support automation, with customers expanding aggressively after successful initial deployments. The approach enables AI agents to operate effectively within specific business contexts without the need for expensive forward-deployed engineers or process redocumentation.
Allen Institute (AI2)
·
4 months ago
Researchers at AI2 released MolmoPoint, a vision-language model architecture that uses special grounding tokens to enable pointing at locations in images and videos instead of generating text coordinates. MolmoPoint-8B achieves 70.7% accuracy on PointBench and 89.2 F1 on PixMo-Points, with three model variants released alongside new datasets including MolmoPoint-GUISyn containing 36,000 synthetic screenshots with 2 million annotated points. The grounding token approach reduces pointing from 8 tokens to 3 tokens, trains faster, and maintains accuracy across different image resolutions better than coordinate-based methods.
Together AI
·
4 months ago
● 2 sources
Together AI expanded its fine-tuning service to support tool calling, reasoning, and vision-language model training, while upgrading infrastructure to handle models up to 1 trillion parameters. The company achieved up to 6× higher throughput for 100B+ parameter models and now supports datasets up to 100GB in size. Users can now iterate faster with improved cost predictions and time estimates before launching training jobs.
