- General data annotation (text/image/video/audio)
- RLHF
- Complex reasoning data production and annotation
- Specialized professional knowledge production and annotation
- Agent benchmark assessment production and annotation
- Multi-language and cross-cultural content production/collection/annotation
- Crowdsourcing evaluation
- ...more
- Tech Enterprises: AI LLM enterprises, Internet platform enterprises, AI Agent enterprises, Embodied robot enterprises, ...
- Professional Service Institutions: Healthcare, Law, Finance, ...
- Government agencies and public institutions
For the common problem of AI LLM hallucination, the client needs to conduct hallucination identification and root cause analysis based on multi-dimensional x-check of generated content and reference content.
Refine hallucination categories:
- Ex nihilo creation
- Date confusion
- Incorrect numerical reference
- Factual misreference
- Logical misinference
Produce and annotate hallucination reasoning chain data:
- Annotate hallucinations and generate direct/indirect inferential evidence
- Annotate contradictions between reference content
- Annotate consistent parts between real content and reference content
- Produce and annotate reasons of no-need-to-judge hallucinations and difficult-to-judge hallucinations
Client's objective: For the accuracy problem of AI LLM in healthcare industry, client needs medical students/doctors with healthcare knowledge and clinical experience to assess and revise the accuracy of the generated content related to specific healthcare issues (including digestive diseases, metabolic diseases, respiratory diseases, etc.)
Highest quality
- Healthcare professional users annotate and revise the errors in the generated content, and additionally supplement healthcare knowledge from the latest academic findings and clinical practice.
- Users with strong self-driven motivation can conduct more precise annotations, broader thinking and judgment dimensions, and create higher quality outputs.
Client's objective: Produce coherent videos based on novel scripts
Full-stack output
- Front-end storyboarding: Users with the abilities in writing film scripts and designing storyboarding visuals produce storyboard scripts by utilizing various tools.
- Mid-end video-creation: Users with abilities in video production and editing, voice-over, and music production create videos based on the storyboard script by using various tools.
- Back-end review: Users with abilities in video editing and content review modify videos.
Shanghai Office: Room 710, Building A, Dongyin Center, Shanghai, China
Beijing Office: Yancheng Building, No.5 Lizheng Road, Beijing, China