MENU
Skip to main content

DigitVibe blog

Generative AI, organizations & the changing world.

Generative AI is changing how we work, build organizations and solve problems. In this blog, AI agents help us find and summarize research insights and discoveries and write about the possibilities they open up.

Subscribe via RSS
  1. Complete abstract from MT-Eval: A Multi-Turn Capabilities Evaluation Benchmark for Large Language Models.
    Generative AI in organizations and society

    Multi-turn AI chats expose weaknesses hidden by single-turn tests

    MT-Eval compares single- and multi-turn conversations and finds that models often lose ground as dialogue history grows.

    Read the article
  2. Complete abstract from Is this the real life? Is this just fantasy? The Misleading Success of Simulating Social Interactions With LLMs.
    AI and social simulation

    Is this the real life? Is this just fantasy? The Misleading Success of Simulating Social Interactions With LLMs

    Simulations with full access to every character’s goals look more successful than interactions where each agent has realistic private information.

    Read the article
  3. Complete abstract from Lookback Lens: Detecting and Mitigating Contextual Hallucinations in Large Language Models Using Only Attention Maps.
    Generative AI reliability

    Can attention maps reveal language-model hallucinations?

    Lookback Lens uses model attention to flag answers that drift from supplied context, with tests across summarization and question answering.

    Read the article
  4. Complete abstract from Do LLMs Plan Like Human Writers? Comparing Journalist Coverage of Press Releases with LLMs.
    Generative AI in organizations and society

    LLMs suggest news angles, but lack human creativity

    A study of journalism planning finds that language models align more closely with reporters on story angles than on sources, while producing less creative suggestions.

    Read the article
  5. Complete abstract from Large Language Models for Data Annotation and Synthesis: A Survey.
    AI and data work

    LLMs Can Help Build Training Data, With Quality Checks

    A survey maps how language models can create labels and examples, how teams assess them, and where synthetic data can introduce new risks.

    Read the article
  6. ReviewCritique Table 1: corpus statistics comparing human-written and LLM-generated reviews, including accepted and rejected paper groups.
    Science and research

    LLMs Assist NLP Researchers: Critique Paper (Meta-)Reviewing

    Experts marked more passages as deficient in AI-written peer reviews than in human reviews. Model feedback was less specific to individual papers.

    Read the article
  7. Complete abstract from The LLM Effect: Are Humans Truly Using LLMs, or Are They Being Influenced By Them Instead?.
    AI and expert work

    The LLM Effect: Are Humans Truly Using LLMs, or Are They Being Influenced By Them Instead?

    A small study of policy analysts finds that LLM suggestions speed topic assignment while increasing alignment with the suggestions themselves.

    Read the article
  8. Complete abstract from LitSearch: A Retrieval Benchmark for Scientific Literature Search.
    Science and research

    LitSearch: Benchmarking AI retrieval of scientific literature

    On 597 literature-search questions, an AI retriever outperformed keyword search. Using GPT-4o to reorder results improved retrieval further.

    Read the article
  9. Complete abstract from Learn to Refuse: Making Large Language Models More Controllable and Reliable through Knowledge Scope Limitation and Refusal Mechanism.
    Generative AI reliability

    When should an AI answer or refuse?

    Learn to Refuse pairs a structured knowledge base with a refusal mechanism and tests whether selective answers improve accuracy.

    Read the article
  10. Complete abstract from LawBench: Benchmarking Legal Knowledge of Large Language Models.
    AI governance and society

    What legal knowledge can language models show?

    LawBench evaluates 21 systems across 20 Chinese legal tasks, revealing uneven skills and weak performance on many practical forms of legal reasoning.

    Read the article