SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2603.06552 · 2026

KCLarity at SemEval-2026 Task 6: Encoder and Zero-Shot Approaches to Political Evasion Detection

Salvatore Greco, Archie Sage

arXiv · PDF · Open in the Atlas

Code that ran

9 of 10 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (0 of them constructed an object rather than computing a result; 1 honoured the contract we drafted, 0 violated it and 1 has no contract verdict; 7 failures of our instrument; 1 no recorded run), measured 2026-11-22; bodies inline only under a permissive licence as GitHub detects it. Not a reproduction of the paper's results.

"Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results. This record, dated, as JSON: /api/ran/2603.06552.json · badge: /api/ran/2603.06552.svg

Each repository's role is a label: a classifier's output recorded on the link, or, where no role is recorded, what the Papers with Code archive marked. The line under it names its basis. Neither says who wrote the code.

RepositoryRoleRan
semeval-2026-kclarity/clarity canonical (classifier label)
classifier basis: census
9 ran · 1 no recorded run
FunctionStatusWhere it lives
add_clarity_special_tokens Ran
ran on a synthesized input
semeval-2026-kclarity/clarity/src/utils/tokenizer_specials.py
code served (permissive licence) · get_code("5a0d7ab9956292e1")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn
from transformers import AutoTokenizer

torch.manual_seed(0)

# Create a minimal tokenizer stand-in
class MinimalTokenizer:
    def __init__(self):
        self.vocab = {
            'hello': 0,
            'world': 1,
            '[CLS]': 2,
            '[SEP]': 3,
        }
        self.special_tokens = set()
    
    def get_vocab(self):
        return self.vocab
    
    def add_special_tokens(self, special_tokens_dict):
        tokens_to_add = special_tokens_dict.get('additional_special_tokens', [])
        count = 0
        for token in tokens_to_add:
            if token not in self.vocab:
                self.vocab[token] = len(self.vocab)
                self.special_tokens.add(token)
                count += 1
        return count
    
    def __len__(self):
        return len(self.vocab)

# Create a minimal model with resize_token_embeddings
class MinimalModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.embedding = nn.Embedding(10, 4)
    
    def resize_token_embeddings(self, new_size):
        old_size = self.embedding.num_embeddings
        self.embedding = nn.Embedding(new_size, 4)
        return self.embedding

# Prepare tokenizer and model
tokenizer = MinimalTokenizer()
model = MinimalModel()

ARGS = {
    'tokenizer': tokenizer,
    'model': model,
    'max_person_tags': 32,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch, transformers — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple — a tuple output has no array shape of its own; none is stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
build_clarity_special_tokens Ran
ran; our contract draft was wrong, not the code
semeval-2026-kclarity/clarity/src/utils/tokenizer_specials.py
code served (permissive licence) · get_code("db6aa52ba81cf129")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned list, ndim 1, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn
from typing import List, Tuple

torch.manual_seed(0)

ARGS = {
    "max_person_tags": 32
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list, ndim 1, dtype family float — a list output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
build_wandb_group_name Ran
ran; our contract draft was wrong, not the code
semeval-2026-kclarity/clarity/src/utils/wandb.py
code served (permissive licence) · get_code("523f9eab0e085c2a")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned str. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np

# Define experiment_config
experiment_config = {
    'experiment_type': 'classification',
    'task_name': 'sentiment',
    'input_format': 'pair',
    'preprocessing': {
        'masking_mode': 'random'
    },
    'class_weights': {
        'use_weighted_loss': True,
        'method': 'balanced',
        'warmup_epochs': 5
    },
    'seed_control': {
        'mode': 'multi'
    },
    'cd': False,
    'hierarchical': {
        'enabled': False,
        'stage': 1
    }
}

# Define task_config
task_config = {
    'task_name': 'sentiment_analysis'
}

# Define model_config
model_config = {
    'model_name': 'bert_base',
    'init_from_checkpoint': False
}

# Define data_config
data_config = {
    'splitting': {
        'method': 'stratified_kfold'
    }
}

ARGS = {
    'experiment_config': experiment_config,
    'task_config': task_config,
    'model_config': model_config,
    'data_config': data_config
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
build_wandb_run_name Ran
ran; our contract draft was wrong, not the code
semeval-2026-kclarity/clarity/src/utils/wandb.py
code served (permissive licence) · get_code("a73236c09044be0f")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned str. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np

# Construct the experiment_config dict
experiment_config = {
    'experiment_name': 'my_ml_experiment',
    'learning_rate': 0.001,
    'batch_size': 32,
    'epochs': 10
}

# Define arguments for build_wandb_run_name
ARGS = {
    'experiment_config': experiment_config,
    'seed': 42,
    'suffix': 'v1'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
compute_basic_metrics Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
semeval-2026-kclarity/clarity/src/utils/metrics.py
code served (permissive licence) · get_code("f8ffd1c560bd2b9b")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned dict. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np

# Set random seed for reproducibility
np.random.seed(0)

# Create small sample data for testing
batch_size = 2
num_classes = 4

# Create predictions as logits (will be argmax-ed by the function)
predictions = np.random.randn(batch_size, num_classes).astype(np.float32)

# Create true labels (class IDs from 0 to num_classes-1)
labels = np.array([0, 2], dtype=np.int64)

ARGS = {
    'predictions': predictions,
    'labels': labels,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, sklearn — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
compute_multi_annotator_accuracy Ran
ran, honoured the contract we drafted
semeval-2026-kclarity/clarity/src/utils/metrics.py
code served (permissive licence) · get_code("343b311ebd371c9d")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned float, ndim 0, dtype family float, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np

# Set random seed for reproducibility
np.random.seed(0)

# Create predictions: array of predicted class IDs for 5 samples
# Classes are in range 0-3 (4 possible classes)
predictions = np.array([0, 1, 2, 3, 1], dtype=np.int64)

# Create annotator labels: 3 annotators, each with labels for 5 samples
# Each annotator has their own labeling of the same 5 samples
annotator_1 = np.array([0, 1, 2, 3, 0], dtype=np.int64)
annotator_2 = np.array([0, 2, 2, 2, 1], dtype=np.int64)
annotator_3 = np.array([1, 1, 2, 3, 1], dtype=np.int64)

annotator_labels = [annotator_1, annotator_2, annotator_3]

ARGS = {
    "predictions": predictions,
    "annotator_labels": annotator_labels,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type float, ndim 0, dtype family float, shape () — shape () as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
compute_multi_annotator_metrics Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
semeval-2026-kclarity/clarity/src/utils/metrics.py
code served (permissive licence) · get_code("b9ce002391bcfa10")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned dict. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np
from typing import Dict, Any, List

np.random.seed(0)

# Create small test data for multi-annotator evaluation
n_samples = 5
n_classes = 3

# Predictions: class IDs for each sample
predictions = np.array([0, 1, 2, 1, 0], dtype=np.int64)

# Three annotators, each providing labels for all samples
annotator_1 = np.array([0, 1, 2, 1, 0], dtype=np.int64)
annotator_2 = np.array([0, 1, 1, 1, 0], dtype=np.int64)
annotator_3 = np.array([1, 1, 2, 2, 0], dtype=np.int64)

annotator_labels = [annotator_1, annotator_2, annotator_3]

# Label names for the classes
label_names = ['class_A', 'class_B', 'class_C']

# Construct ARGS dictionary
ARGS = {
    'predictions': predictions,
    'annotator_labels': annotator_labels,
    'label_names': label_names
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, sklearn — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
load_experiment_config Ran
ran; our contract draft was wrong, not the code
semeval-2026-kclarity/clarity/src/utils/config.py
code served (permissive licence) · get_code("68a71a0ae43a6b96")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned dict. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import os
import tempfile
import yaml

# Create temporary directory for config files
temp_dir = tempfile.mkdtemp()

# Create task config
task_config_path = os.path.join(temp_dir, 'task.yaml')
task_config_content = {
    'task_name': 'classification',
    'num_classes': 10,
    'metrics': ['accuracy', 'f1']
}
with open(task_config_path, 'w') as f:
    yaml.dump(task_config_content, f)

# Create model config
model_config_path = os.path.join(temp_dir, 'model.yaml')
model_config_content = {
    'model_type': 'transformer',
    'hidden_size': 256,
    'num_layers': 2,
    'model_name': '{experiment_name}_model'
}
with open(model_config_path, 'w') as f:
    yaml.dump(model_config_content, f)

# Create data config
data_config_path = os.path.join(temp_dir, 'data.yaml')
data_config_content = {
    'dataset_name': 'mnist',
    'batch_size': 32,
    'data_dir': '/tmp/data_{task_name}'
}
with open(data_config_path, 'w') as f:
    yaml.dump(data_config_content, f)

# Create experiment config
experiment_config_path = os.path.join(temp_dir, 'experiment.yaml')
experiment_config_content = {
    'experiment_name': 'exp_001',
    'experiment_type': 'baseline',
    'model_config': model_config_path,
    'task_config': task_config_path,
    'data_config': data_config_path,
    'training': {
        'learning_rate': '0.001',
        'warmup_ratio': '0.1',
        'weight_decay': '1e-5',
        'num_epochs': 10
    }
}
with open(experiment_config_path, 'w') as f:
    yaml.dump(experiment_config_content, f)

ARGS = {
    'experiment_config_path': experiment_config_path
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
yaml — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
substitute_variables Ran
ran; our contract draft was wrong, not the code
semeval-2026-kclarity/clarity/src/utils/config.py
code served (permissive licence) · get_code("56434a98e100ccbe")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned dict. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np

torch.manual_seed(0)

# Construct arguments for substitute_variables function

config = {
    "experiment_name": "test_{experiment_name}",
    "experiment_type": "{experiment_type}_baseline",
    "task_name": "task_{task_name}",
    "nested": {
        "path": "/data/{experiment_name}/{task_name}",
        "value": 42
    },
    "list_values": [
        "prefix_{experiment_name}",
        "middle_{experiment_type}",
        "suffix"
    ],
    "mixed": {
        "deep": {
            "setting": "{task_name}_v2"
        },
        "number": 100
    }
}

variables = {
    "experiment_name": "exp_001",
    "experiment_type": "supervised",
    "task_name": "classification"
}

ARGS = {
    "config": config,
    "variables": variables
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
load_config Not yet run
unverified (harvested, no recorded run)
semeval-2026-kclarity/clarity/src/utils/config.py
code served (permissive licence) · get_code("b7c98f44a42d7c36")
How it ran

no invocation recorded.

Also present in other papers' code

Of 10 samples lifted from this paper's linked repositories, 10 carry a function identity and 0 do not. Of the 10 distinct functions with an identity, 0 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 0 later) and 10 are only here among the 52,977 papers whose code Syntology holds (as of 2026-09-26).

Only here means: among the 52,977 papers whose code Syntology holds (51,887 with lifted samples, 1,090 through their assemblies only), as of 2026-09-26, no other paper's lifted samples or current assembly bricks contain an identical function body after normalisation.

Presence, not direction: an identical function body in two papers' repositories does not say who wrote it, which came from which, or that either paper's method rests on it. Matching is exact after normalisation, so a renamed or edited function is not matched. Dates are when papers appeared, not when code was written. Only .py files the harvester lifted are covered; notebooks are not read. A new-style arXiv id YYMM.NNNN dates a paper to the month 20YY-MM; otherwise Paper.year dates it to the year; otherwise it is undated. Dates are when the paper appeared, not when the code was written. Read 2026-11-22.

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

This paper describes the KCLarity team's participation in CLARITY, a shared task at Se-mEval 2026 on classifying ambiguity and evasion techniques in political discourse. We investigate two modelling formulations: (i) directly predicting the clarity label, and (ii) predicting the evasion label and deriving clarity through the task taxonomy hierarchy. We further explore several auxiliary training variants and evaluate decoder-only models in a zeroshot setting under the evasion-first formulation. Overall, the two formulations yield comparable performance. Among encoder-based models, RoBERTa-large achieves the strongest results on the public test set, while zero-shot GPT-5.2 generalises better on the hidden evaluation set.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2603.06552")
get_code_for_paper("2603.06552")
have("2603.06552")

The run record, dated, one paper per request, free:

curl https://syntology.ai/api/ran/2603.06552.json

A badge for a README (the split and the date, never a ratio):

[![Syntology run record](https://syntology.ai/api/ran/2603.06552.svg)](https://syntology.ai/paper/2603.06552)

Connect an agent — have() is free.