Skip to content

Paper Group: Explicit Modeling of Uncertainty with an [IDK] Token

Photo of Logan
Hosted By
Logan

Details

Join us for a paper discussion on "I Don't Know: Explicit Modeling of Uncertainty with an [IDK] Token"

  • Analyzing hallucination reduction through dedicated uncertainty tokens in language models
    Featured Paper:
    "I Don't Know: Explicit Modeling of Uncertainty with an [IDK] Token" (Author et al., 2024)
    arXiv Paper
    Performance Benchmarks
    | Metric | [IDK] Models | Baseline Models | Improvement |
    | ------ | ------------ | --------------- | ----------- |
    | Hallucination Rate | 12.4% | 23.7% | -47.7% |
    | Knowledge Retention | 94.1% | 95.3% | -1.2% |
    | Abstention Accuracy | 88.6% | N/A | New Metric |
    Implementation Challenges
  • Probability mass redistribution during inference
  • Temperature scaling for uncertainty calibration
  • Compatibility with existing RLHF pipelines

Key Technical Features

  • 0.03% vocabulary size increase (1 new token)
  • 15% training time overhead vs standard fine-tuning
  • Linear probe analysis of uncertainty patterns

Future Directions

  • Multilingual [IDK] token alignment
  • Extension to multimodal uncertainty signaling
  • Integration with constitutional AI frameworks

Silicon Valley Generative AI has two meeting formats:

  • Paper Reading -
    Every second week we meet to discuss machine learning papers. This is a collaboration between Silicon Valley Generative AI and Boulder Data Science.Talks - Once a month we meet to have someone present on a topic related to generative AI. Speakers can range from industry leaders, researchers, startup founders, subject matter experts and those with an interest in a topic and would like to share. Topics vary from technical to business focused. They can be on how the latest in generative models work and how they can be used, applications and adoption of generative AI, demos of projects and startup pitches or legal and ethical topics. The talks are meant to be inclusive and for a more general audience compared to the paper readings.

If you would like to be a speaker or suggest a paper email us @ svb.ai.paper.suggestions@gmail.com or join our new discord !!!

Photo of Boulder Data Science, Machine Learning & AI group
Boulder Data Science, Machine Learning & AI
See more events
FREE