Auto71%Harbor: pendingKnowledge & Reasoning

VaccineStance

A human-annotated benchmark for vaccine stance detection constructed from social media posts collected between 2013 and 2022.

How to run

ltb · vaccinestance
$ npm i -g lettokenburn-cli
$ ltb bench info vaccinestance

External harness. No Harbor adapter has been generated for this benchmark yet. `ltb bench info` prints the known links.

Runs happen on your machine with your own subscription; only the score summary is pushed. CLI docs

Scores

5 sourced entries
HarnessSourceConfidenceLink
Llama60.2N/AIndependentmediumOct 9, 2026arxiv.org
GPT-OSS58.8N/AIndependentmediumOct 9, 2026arxiv.org
Mistral55.6N/AIndependentmediumOct 9, 2026arxiv.org
Phi49.5N/AIndependentmediumOct 9, 2026arxiv.org
Gemma37.8N/AIndependentmediumOct 9, 2026arxiv.org

Every number links to where it was published. Low-confidence rows are muted and need per-cell verification.

Reproduced by the community

All runs
Loading community runs…

About

VaccineStance consists of human-annotated social media posts related to vaccines, categorized into In-Favor, Against, or Neutral stances. It was constructed from data collected between 2013 and 2022.

The benchmark is evaluated using Macro-F1 score. Experiments show that the proposed MoPrO-SD framework improves upon human-written base prompts across various LLMs and settings.