AI Paper Atlas

Research map / Probabilistic and scientific modeling / Variational and generative sampling

Estimating Tail Risks in Language Model Output Distributions

Rico Angell, Raghav Singhal, Zachary Horvitz, Zhou Yu, Rajesh Ranganath, Kathleen McKeown, He He

ICML 2026 · Probabilistic and scientific modeling › Variational and generative sampling · approach: Sequential Monte Carlo for inference · 0 citations (OpenAlex)

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 changedWhereMeasuredEffect
sample 10,000 outputs per query instead of one greedy outputsafety-tuned instruct models · STRONGREJECT 313 queries% queries with at least one harmful output: Llama-3.1-8B: 1.6% -> 74.0%; Llama-3.2-1B: 16.6% -> 99.5%Improved
estimate harm probability with importance sampling on a steered unsafe proposal instead of naive Monte Carlotarget LLMs (Llama, Qwen, Olmo, Phi-4) · STRONGREJECTestimation error vs 10K-sample MC: matches brute force with 10-20x fewer samples (k=1000 vs 10,000)Improved
use importance sampling for sycophancy and hallucination rate estimationQwen2.5-7B and Llama-3.1-8B · 20 safe leading queries per traitestimation errorImproved
optimize proposal hyperparameters (steering intensity, switch index, mixing coefficient) with cross-entropy objective (ablation)steered proposal model · STRONGREJECT, k=500 samplesmean squared error vs brute forceImproved
rewrite queries with non-adversarial paraphrasesLlama-3.1-8B, Phi-4, Qwen2.5-7B · 309 STRONGREJECT queries, 25 rewrites eachharm probability: Qwen2.5-7B: 0.0 -> 0.14 for one queryHurt
forecast worst-case risk on unseen queries using extreme value theory from evaluation-set QRisk estimatesquery-level estimates · STRONGREJECT paraphrases, 30% eval / 70% deploymentprobability of max query risk above thresholdImproved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy