Тест, свързан с ръководство · Среден Ниво

Викторина на института Алън за AI

Разгледайте мисията на AI2, потока от отворени модели на OLMo, след обучението на OLMo 3, Dolma, AllenNLP и API на Academic Graph на Semantic Scholar.

Свързани ръководни пътекиИнститут Алън за AI
Въпрос 1 на 8

The Allen Institute for AI (AI2) was founded by whom?

Визуализирайте всеки въпрос8 въпроси

Отговорите и обясненията се появяват, докато отговаряте на теста.

  1. The Allen Institute for AI (AI2) was founded by whom?
    • Пол Алън
    • Бил Гейтс
    • Илон Мъск
    • Сам Алтман
  2. What does AI2 mean by OLMo’s fully open model flow?
    • Only final weights are released
    • It provides access to lifecycle artifacts such as data, code, checkpoints, and reports, not only final weights
    • Every use of every artifact is unrestricted
    • The model exists only as a paid API
  3. Which AI2 service supports scholarly search and discovery?
    • AllenNLP
    • Molmo
    • Semantic Scholar
    • Dolma
  4. How does Ai2 state its current mission on its About page?
    • AI за общото благо
    • Replacing every research institution with one model
    • Building only commercial chatbots
    • Building breakthrough AI to solve the world’s biggest problems
  5. What is AllenNLP, and what is the status of its official repository?
    • An open-source NLP research library built on PyTorch; its repository is archived and read-only
    • A current hosted language model with active weight releases
    • A scholarly search service operated by Semantic Scholar
    • A commercial data warehouse for AI experiments
  6. Which statement correctly scopes the original Dolma release?
    • Every OLMo model used exactly 3 trillion tokens
    • The 2023 corpus was described as 3 trillion tokens; later releases such as Dolma 3 differ
    • Dolma is a court-opinion index
    • The corpus contains only code
  7. Which post-training stages does AI2 list for OLMo 3?
    • Only image-caption training
    • Only masked-language pretraining
    • Supervised fine-tuning, preference tuning with DPO, and reinforcement learning with verifiable rewards
    • No post-training stages
  8. What type of records are included in the Semantic Scholar Academic Graph (S2AG)?
    • Only language-model weights
    • Only court contracts
    • Only satellite imagery
    • Scholarly paper titles, abstracts, citations, and related metadata