CLARITY: Unmasking Political Question Evasions
SemEval-2026 Task 6 (CLARITY): detecting how directly a politician answers a question. Given a question-answer pair from a U.S. presidential interview, annotators classify the reply into one of three clarity levels (Clear Reply, Ambiguous, Clear Non-Reply) and, when the answer is evasive, identify the evasion technique used. Multi-part questions are first decomposed into single sub-questions so each judgment captures how well one specific inquiry is addressed.
Configuration Fileconfig.yaml
This Potato config reproduces the annotation task. Save it as config.yaml and run potato start config.yaml to try it.
# CLARITY: Unmasking Political Question Evasions
# Based on Thomas, Filandrianos, Lymperaiou, Zerva, and Stamou, SemEval-2026 Task 6
# Paper: https://aclanthology.org/2026.semeval-1.449/
# Task page: https://konstantinosftw.github.io/CLARITY-SemEval-2026/
#
# Given a question-answer (QA) pair from a U.S. presidential interview
# (2006-2023), judge how directly the answer addresses the question. The
# task has two nested subtasks:
# - Subtask 1 (Clarity): 3 coarse clarity tiers
# - Subtask 2 (Evasion): one of 9 fine-grained evasion techniques when
# the answer is not a clear reply
# Multi-part questions are decomposed into single sub-questions so each
# annotation captures how well one specific inquiry is addressed.
#
# Subtask 1 clarity labels (from the official task page):
# - Clear Reply: the answer directly and unambiguously addresses the question
# - Ambiguous: the answer partially addresses it or is unclear
# - Clear Non-Reply: the answer avoids the question
#
# Subtask 2 uses a taxonomy of 9 evasion techniques (e.g. deflection,
# partial answer, declining to answer, attacking the question). This
# showcase collects the technique as free text; consult the task paper for
# the full 9-way label set.
annotation_task_name: "CLARITY - Political Question Evasion"
task_dir: "."
data_files:
- sample-data.json
item_properties:
id_key: "id"
text_key: "answer"
output_annotation_dir: "annotation_output/"
output_annotation_format: "json"
port: 8000
server_name: localhost
annotation_schemes:
- annotation_type: radio
name: clarity
description: "How directly does the answer address the question?"
labels:
- "Clear Reply"
- "Ambiguous"
- "Clear Non-Reply"
keyboard_shortcuts:
"Clear Reply": "1"
"Ambiguous": "2"
"Clear Non-Reply": "3"
tooltips:
"Clear Reply": "The answer directly and unambiguously addresses the question"
"Ambiguous": "The answer only partially addresses the question or is unclear"
"Clear Non-Reply": "The answer avoids or does not address the question"
- annotation_type: text
name: evasion_technique
description: "If the answer is not a Clear Reply, which evasion technique is used? (see task taxonomy of 9 techniques)"
textarea: false
required: false
placeholder: "e.g. deflection, partial answer, declining to answer..."
annotation_instructions: |
Read the question and the politician's answer. Decide whether the answer
is a Clear Reply, is Ambiguous, or is a Clear Non-Reply to that specific
question. If it is not a Clear Reply, note which evasion technique the
respondent used.
html_layout: |
<div style="padding: 15px; max-width: 820px; margin: auto;">
<div style="background: #eff6ff; border: 1px solid #bfdbfe; border-radius: 8px; padding: 14px; margin-bottom: 10px;">
<strong style="color: #1e40af;">Question:</strong>
<p style="font-size: 16px; line-height: 1.6; margin: 6px 0 0 0;">{{question}}</p>
</div>
<div style="background: #f8fafc; border: 1px solid #e2e8f0; border-radius: 8px; padding: 14px;">
<strong style="color: #334155;">Answer:</strong>
<p style="font-size: 16px; line-height: 1.6; margin: 6px 0 0 0;">{{answer}}</p>
</div>
</div>
allow_all_users: true
instances_per_annotator: 50
annotation_per_instance: 3
allow_skip: true
Sample Datasample-data.json
[
{
"id": "clr_001",
"question": "Will you raise taxes on the middle class next year?",
"answer": "No, I have said repeatedly that no family earning under $400,000 will see a tax increase."
},
{
"id": "clr_002",
"question": "Do you support a ceasefire in the conflict?",
"answer": "What we all want is peace, and I think everyone in this country wants to see stability in the region."
}
]
// ... and 8 more itemsGet This Design
Clone or download from the repository
Quick start:
git clone https://github.com/davidjurgens/potato-showcase.git cd potato-showcase/semeval/2026/task06-clarity-political-evasion potato start config.yaml
Dataset & paper
Thomas et al., SemEval 2026 (Task 6)
Citation (BibTeX)
@inproceedings{thomas-etal-2026-semeval,
title = "{S}em{E}val-2026 Task 6: {CLARITY} -- Unmasking Political Question Evasions",
author = "Thomas, Konstantinos and Filandrianos, Giorgos and Lymperaiou, Maria and Zerva, Chrysoula and Stamou, Giorgos",
booktitle = "Proceedings of the 20th International Workshop on Semantic Evaluation (SemEval-2026)",
year = "2026",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.semeval-1.449",
pages = "3704--3715"
}Details
Annotation Types
Domain
Use Cases
Tags
Found an issue or want to improve this design?
Open an IssueRelated Designs
BRAINTEASER - Commonsense-Defying QA
Lateral thinking and commonsense-defying question answering task requiring annotators to select answers to brain teasers that defy default commonsense assumptions and provide explanations. Based on SemEval-2024 Task 9 (BRAINTEASER).
Clickbait Spoiling
Classification and extraction of spoilers for clickbait posts, including spoiler type identification and span-level spoiler detection. Based on SemEval-2023 Task 5 (Fröbe et al.).
Clinical Trial NLI
Natural language inference for clinical trial reports, determining whether a given statement is entailed or contradicted by clinical trial evidence. Based on SemEval-2023 Task 7 (Jullien et al.).