SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2108.01879 · EMNLP · 2021

SUMMARY EXPLORER Visualizing the State of the Art in Text Summarization

Martin Potthast, Khalid Al-Khatib, Shahbaz Syed, Tariq Yousef, Stefan Jänicke

arXiv · PDF · Open in the Atlas

Code that ran

11 of 15 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (1 of them constructed an object rather than computing a result; 0 honoured the contract we drafted, 0 violated it and 6 have no contract verdict; 5 failures of our instrument; 4 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/2108.01879.json · badge: /api/ran/2108.01879.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
webis-de/summary-explorer canonical (classifier label)
classifier basis: census
5 ran · 4 no recorded run
pltrdy/rouge community
Papers with Code archive
6 ran · 0 no recorded run
FunctionStatusWhere it lives
Ngrams Ran
ran on a synthesized input · constructed an object; no result was computed
pltrdy/rouge/rouge/rouge_score.py
code served (permissive licence) · get_code("0b1b6496ee19d475")
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 an instance of its own class, Ngrams: the run constructed the object and did not call it, so no result was computed. 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
from copy import deepcopy

torch.manual_seed(0)

# The Ngrams class constructor takes:
# - ngrams: a collection (set or list) of ngram objects
# - exclusive: boolean flag

# Looking at how the class is used:
# - ngrams parameter is passed to set() or list() constructor
# - Methods like intersection() and union() work with ._ngrams attribute
# - The class stores tuples or hashable objects as ngrams

# Create sample ngrams (using tuples as they're hashable and work with both set/list)
ngrams_data_exclusive = {("the", "cat"), ("cat", "sat"), ("sat", "on")}
ngrams_data_non_exclusive = [("the", "cat"), ("cat", "sat"), ("sat", "on"), ("cat", "sat")]

ARGS = {
    "ngrams": ngrams_data_exclusive,
    "exclusive": True
}
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 Ngrams — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a Ngrams 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
_get_ngrams Ran
ran; our contract draft was wrong, not the code
pltrdy/rouge/rouge/rouge_score.py
code served (permissive licence) · get_code("699442d33ab83bd2")
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 Ngrams. 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
from copy import deepcopy

class Ngrams(object):
    """
        Ngrams datastructure based on `set` or `list`
        depending in `exclusive`
    """

    def __init__(self, ngrams={}, exclusive=True):
        if exclusive:
            self._ngrams = set(ngrams)
        else:
            self._ngrams = list(ngrams)
        self.exclusive = exclusive

    def add(self, o):
        if self.exclusive:
            self._ngrams.add(o)
        else:
            self._ngrams.append(o)

    def __len__(self):
        return len(self._ngrams)

    def intersection(self, o):
        if self.exclusive:
            inter_set = self._ngrams.intersection(o._ngrams)
            return Ngrams(inter_set, exclusive=True)
        else:
            other_list = deepcopy(o._ngrams)
            inter_list = []
            for e in self._ngrams:
                try:
                    i = other_list.index(e)
                except ValueError:
                    continue
                other_list.pop(i)
                inter_list.append(e)
            return Ngrams(inter_list, exclusive=False)

    def union(self, *ngrams):
        if self.exclusive:
            union_set = self._ngrams
            for o in ngrams:
                union_set = union_set.union(o._ngrams)
            return Ngrams(union_set, exclusive=True)
        else:
            union_list = deepcopy(self._ngrams)
            for o in ngrams:
                union_list.extend(o._ngrams)
            return Ngrams(union_list, exclusive=False)

def _get_ngrams(n, text, exclusive=True):
    """Calcualtes n-grams.

    Args:
      n: which n-grams to calculate
      text: An array of tokens

    Returns:
      A set of n-grams
    """
    ngram_set = Ngrams(exclusive=exclusive)
    text_length = len(text)
    max_index_ngram_start = text_length - n
    for i in range(max_index_ngram_start + 1):
        ngram_set.add(tuple(text[i:i + n]))
    return ngram_set

ARGS = {
    "n": 2,
    "text": ["hello", "world", "foo", "bar"],
    "exclusive": True
}
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 Ngrams — a Ngrams output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Ngrams 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
_get_word_ngrams Ran
ran; our contract draft was wrong, not the code
pltrdy/rouge/rouge/rouge_score.py
code served (permissive licence) · get_code("82942ade3a066704")
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 Ngrams. 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 itertools
from copy import deepcopy

class Ngrams(object):
    """
        Ngrams datastructure based on `set` or `list`
        depending in `exclusive`
    """

    def __init__(self, ngrams={}, exclusive=True):
        if exclusive:
            self._ngrams = set(ngrams)
        else:
            self._ngrams = list(ngrams)
        self.exclusive = exclusive

    def add(self, o):
        if self.exclusive:
            self._ngrams.add(o)
        else:
            self._ngrams.append(o)

    def __len__(self):
        return len(self._ngrams)

    def intersection(self, o):
        if self.exclusive:
            inter_set = self._ngrams.intersection(o._ngrams)
            return Ngrams(inter_set, exclusive=True)
        else:
            other_list = deepcopy(o._ngrams)
            inter_list = []
            for e in self._ngrams:
                try:
                    i = other_list.index(e)
                except ValueError:
                    continue
                other_list.pop(i)
                inter_list.append(e)
            return Ngrams(inter_list, exclusive=False)

    def union(self, *ngrams):
        if self.exclusive:
            union_set = self._ngrams
            for o in ngrams:
                union_set = union_set.union(o._ngrams)
            return Ngrams(union_set, exclusive=True)
        else:
            union_list = deepcopy(self._ngrams)
            for o in ngrams:
                union_list.extend(o._ngrams)
            return Ngrams(union_list, exclusive=False)

def _get_ngrams(n, text, exclusive=True):
    """Calcualtes n-grams.

    Args:
      n: which n-grams to calculate
      text: An array of tokens

    Returns:
      A set of n-grams
    """
    ngram_set = Ngrams(exclusive=exclusive)
    text_length = len(text)
    max_index_ngram_start = text_length - n
    for i in range(max_index_ngram_start + 1):
        ngram_set.add(tuple(text[i:i + n]))
    return ngram_set

def _split_into_words(sentences):
    """Splits multiple sentences into words and flattens the result"""
    return list(itertools.chain(*[_.split(' ') for _ in sentences]))

def _get_word_ngrams(n, sentences, exclusive=True):
    """Calculates word n-grams for multiple sentences.
    """
    assert len(sentences) > 0
    assert n > 0
    words = _split_into_words(sentences)
    return _get_ngrams(n, words, exclusive=exclusive)


ARGS = {
    "n": 2,
    "sentences": ["hello world test", "foo bar baz"],
    "exclusive": True
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Ngrams — a Ngrams output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Ngrams 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
_split_into_words Ran
ran; our contract draft was wrong, not the code
pltrdy/rouge/rouge/rouge_score.py
code served (permissive licence) · get_code("6a189ad1c6a91653")
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 itertools

def _split_into_words(sentences):
    """Splits multiple sentences into words and flattens the result"""
    return list(itertools.chain(*[_.split(' ') for _ in sentences]))

ARGS = {
    'sentences': ['hello world', 'foo bar baz', 'test']
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — 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
clean_text Ran
ran on a synthesized input
webis-de/summary-explorer/text-processing/src/utils.py
code served (permissive licence) · get_code("f7918be2e65b31a7")
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

# The function clean_text takes a single parameter 'text' which should be a string
# It performs various string replacements on the input

# Create a sample text with various tokens and patterns that the function handles
sample_text = "This is a test <t>with</t> some -lrb- brackets -rrb- and n't contractions and ' quotes and \" double quotes and , commas and . periods and ( parentheses ) here </t> <n>more</n> -LRB- uppercase -RRB- and <s>tags</s>"

ARGS = {
    "text": sample_text
}
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 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
detokenize_relation Ran
ran on a synthesized input
webis-de/summary-explorer/text-processing/src/utils.py
code served (permissive licence) · get_code("4b2ce6bb44dca364")
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

def detokenize_text(text):
    return text.replace(" n't", "n't").replace(" '", "'").replace(' "', '"').replace(' ,', ',').replace('` ', '`').replace(' .', '.').replace('( ', '(').replace(' )', ')')

def detokenize_relation(relation):
    return {k: detokenize_text(v) for k, v in relation.items()}

# Construct a sample relation dict with tokenized text values
relation = {
    "subject": "The quick brown fox ( jumps )",
    "predicate": "jumps over",
    "object": "the lazy dog .",
    "context": "In a story , there was n't a better moment ."
}

ARGS = {
    "relation": relation
}
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 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
detokenize_text Ran
ran on a synthesized input
webis-de/summary-explorer/text-processing/src/utils.py
code served (permissive licence) · get_code("962bbc77ef68d643")
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

def detokenize_text(text):
    return text.replace(" n't", "n't").replace(" '", "'").replace(' "', '"').replace(' ,', ',').replace('` ', '`').replace(' .', '.').replace('( ', '(').replace(' )', ')')

ARGS = {
    "text": "This is a test . It has ( some ) punctuation , n't and ' quotes with ` backticks ."
}
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 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
f_r_p_rouge_n Ran
ran; our contract draft was wrong, not the code
pltrdy/rouge/rouge/rouge_score.py
code served (permissive licence) · get_code("6f967ee0bce679ea")
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

ARGS = {
    'evaluated_count': 10,
    'reference_count': 8,
    'overlapping_count': 6
}
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 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
get_entity_counts Ran
ran on a synthesized input
webis-de/summary-explorer/text-processing/src/step_1_nlp_pipeline.py
code served (permissive licence) · get_code("a9dd30c56406efb3")
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.

from itertools import chain

# Construct a minimal processed_record that satisfies the function's requirements
# The function expects:
# - processed_record['sentences'] to be a list of dicts
# - each sentence dict to have an 'entities' key
# - each entity to be a tuple/list of (text, type)

processed_record = {
    'sentences': [
        {
            'entities': [
                ('Apple', 'ORG'),
                ('Steve Jobs', 'PERSON'),
            ]
        },
        {
            'entities': [
                ('California', 'LOC'),
                ('Apple', 'ORG'),
            ]
        },
        {
            'entities': []  # Empty entities list to test filtering
        },
        {
            'entities': [
                ('Steve Jobs', 'PERSON'),
            ]
        },
    ]
}

ARGS = {
    'processed_record': processed_record
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — 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
get_remaining_content Ran
ran on a synthesized input
webis-de/summary-explorer/text-processing/src/step_2_lexical_alignment.py
code served (permissive licence) · get_code("3b4c00097055bedb")
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

ARGS = {
    "candidate": "The quick brown fox jumps over the lazy dog",
    "matched": "The quick fox"
}
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 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
rouge_n Ran
ran; our contract draft was wrong, not the code
pltrdy/rouge/rouge/rouge_score.py
code served (permissive licence) · get_code("dd3dfb07f3d3bf48")
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) 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 itertools
from copy import deepcopy

# Define ARGS dict with all parameters for rouge_n function
ARGS = {
    'evaluated_sentences': ['this is a test', 'another sentence here'],
    'reference_sentences': ['this is a test', 'yet another sentence'],
    'n': 2,
    'raw_results': False,
    'exclusive': True
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
none outside the standard library — 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
bert_score_similarity Not yet run
unverified (harvested, no recorded run)
webis-de/summary-explorer/text-processing/src/step_3_semantic_alignment.py
code served (permissive licence) · get_code("a7aa528d3a102206")
How it ran

no invocation recorded.

compute_bert_score Not yet run
unverified (harvested, no recorded run)
webis-de/summary-explorer/text-processing/src/metrics.py
code served (permissive licence) · get_code("71936af5d44993bd")
How it ran

no invocation recorded.

compute_mean_rouge_overlap Not yet run
unverified (harvested, no recorded run)
webis-de/summary-explorer/text-processing/src/metrics.py
code served (permissive licence) · get_code("6cb2e57033eb24ca")
How it ran

no invocation recorded.

compute_rouge_score Not yet run
unverified (harvested, no recorded run)
webis-de/summary-explorer/text-processing/src/metrics.py
code served (permissive licence) · get_code("1b84d2e60b85384e")
How it ran

no invocation recorded.

Also present in other papers' code

Of 15 samples lifted from this paper's linked repositories, 15 carry a function identity and 0 do not. Of the 15 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 15 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 introduces SUMMARY EXPLORER, a new tool to support the manual inspection of text summarization systems by compiling the outputs of 55 state-of-the-art single document summarization approaches on three benchmark datasets, and visually exploring them during a qualitative assessment. The underlying design of the tool considers three well-known summary quality criteria (coverage, faithfulness, and position bias), encapsulated in a guided assessment based on tailored visualizations. The tool complements existing approaches for locally debugging summarization models and improves upon them. The tool is available at https://tldr.webis.de/.

For agents

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

get_harvested_code_for_paper("2108.01879")
get_code_for_paper("2108.01879")
have("2108.01879")

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

curl https://syntology.ai/api/ran/2108.01879.json

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

[![Syntology run record](https://syntology.ai/api/ran/2108.01879.svg)](https://syntology.ai/paper/2108.01879)

Connect an agent — have() is free.