Research map / Large language models / Language, safety and interpretability
Rare Event Analysis of Large Language Models
Jake McAllister Dorman, Edward Gillman, Dominic C Rose, Jamie F. Mair, Juan P. Garrahan
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| estimate rare-completion probabilities with annealed TPS plus MBAR instead of direct sampling at equal token budget | TinyStories-8M · ARI and Log-Prob distribution tails | tail coverage and relative CI width | Improved |
| use MBAR histogram instead of direct histogram for ARI beyond the training data range | TinyStories-8M · ARI | range of ARI values covered | Improved |
| double the number of MCMC steps per chain (ablation) | annealed TPS · ARI and Log-Prob | change in bin heights | Mixed |
| count consecutive token repeats as cheap proxy for high ARI | TinyStories-8M completions · ARI-biased samples | repeat count vs ARI | Improved |
| cap ARI at 15 in the observable (ablation) | TPS sampling · ARI | MCMC acceptance rate | Improved |
Closest papers by meaning
- A Probabilistic Perspective on Unlearning and Alignment for Large Language Models (ICLR 2025 · similarity 0.89)
- Measuring what Matters: Construct Validity in Large Language Model Benchmarks (NeurIPS 2025 · similarity 0.89)
- Systematic Outliers in Large Language Models (ICLR 2025 · similarity 0.88)
- Is a Large Language Model a Good Annotator for Event Extraction? (AAAI 2024 · similarity 0.88)
- Large Language Models Must Be Taught to Know What They Don’t Know (NeurIPS 2024 · similarity 0.87)
- What is Wrong with Perplexity for Long-context Language Modeling? (ICLR 2025 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
