SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2304.08295 · CVPR · 2023

Interactive and Explainable Region-guided Radiology Report Generation

Daniel Rueckert, Georgios Kaissis, Philip Müller, Tim Tanida

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; 0 honoured the contract we drafted, 0 violated it and 9 have no contract verdict; 0 failures of our instrument; 1 no recorded run), measured 2026-11-22; repository links from the Papers with Code archive attributed under CC BY-SA, no affiliation; 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/2304.08295.json · badge: /api/ran/2304.08295.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
ttanida/rgrg canonical (classifier label)
classifier basis: census
9 ran · 1 no recorded run
FunctionStatusWhere it lives
check_coordinate Ran
ran on a synthesized input
ttanida/rgrg/src/dataset/create_dataset.py
code served (permissive licence) · get_code("55ca647b5679ac24")
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 int, ndim 0, dtype family int, 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 torch

torch.manual_seed(0)

ARGS = {
    "coordinate": 5,
    "dim": 10
}
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 int, ndim 0, dtype family int, 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
convert_generated_sentences_to_report Ran
ran on a synthesized input
ttanida/rgrg/src/full_model/generate_reports_for_images.py
code served (permissive licence) · get_code("fe691a9d3026f391")
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
from collections import defaultdict


class MockSentenceTokenizer:
    """Mock sentence tokenizer that mimics spaCy nlp object behavior"""
    
    def __call__(self, text):
        """Tokenize text into sentences"""
        sentences = text.split('. ')
        return MockDoc(sentences)


class MockSent:
    """Mock sentence object"""
    
    def __init__(self, text):
        self.text = text


class MockDoc:
    """Mock document object containing sentences"""
    
    def __init__(self, sentences):
        self.sents = [MockSent(sent.strip()) for sent in sentences if sent.strip()]


class MockBertScore:
    """Mock BertScore compute function"""
    
    def compute(self, lang, predictions, references, model_type):
        """Mock compute method that returns f1 scores"""
        # Return high score if predictions and references are similar enough
        # For testing, return different scores based on input
        if predictions[0] == references[0]:
            f1_score = 0.95  # High similarity for identical sentences
        elif len(predictions[0]) > 50 and len(references[0]) > 50:
            f1_score = 0.92  # Moderate-high similarity for long sentences
        else:
            f1_score = 0.75  # Low similarity for different sentences
        
        return {
            'precision': [f1_score],
            'recall': [f1_score],
            'f1': [f1_score]
        }


# Create mock objects
mock_sentence_tokenizer = MockSentenceTokenizer()
mock_bert_score = MockBertScore()

# Create input for the function
generated_sents_for_selected_regions = [
    "The patient shows normal cardiac function.",
    "No abnormalities detected in the imaging study.",
    "The heart appears normal on examination."
]

ARGS = {
    'generated_sents_for_selected_regions': generated_sents_for_selected_regions,
    'bert_score': mock_bert_score,
    'sentence_tokenizer': mock_sentence_tokenizer
}
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
coordinates_faulty Ran
ran on a synthesized input
ttanida/rgrg/src/dataset/compute_stats_dataset.py
code served (permissive licence) · get_code("68147cc7ae81ca99")
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 bool, 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 torch

# Call coordinates_faulty with valid bounding box arguments
# height: image height
# width: image width
# x1, y1: top-left corner of bbox
# x2, y2: bottom-right corner of bbox

ARGS = {
    "height": 100,
    "width": 150,
    "x1": 10,
    "y1": 20,
    "x2": 50,
    "y2": 80
}
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 bool, 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
determine_if_abnormal Ran
ran on a synthesized input
ttanida/rgrg/src/dataset/compute_stats_dataset.py
code served (permissive licence) · get_code("fd51ccc7bbc208e4")
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 bool, 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 torch

ARGS = {
    "attributes_list": [
        ['anatomicalfinding|no|lung opacity', 'anatomicalfinding|no|pneumothorax', 'nlp|yes|normal'],
        ['anatomicalfinding|no|pneumothorax']
    ]
}
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 bool, 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
get_mean_std Ran
ran on a synthesized input
ttanida/rgrg/src/dataset/compute_mean_std_dataset.py
code served (permissive licence) · get_code("fa575de86a2b3603")
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, 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 cv2
import numpy as np
import tempfile
import os

# Create temporary directory and small test images
temp_dir = tempfile.mkdtemp()

# Create a few small test images as PNG files
image_paths = []
for i in range(3):
    img = np.random.randint(0, 256, (4, 4, 3), dtype=np.uint8)
    img_path = os.path.join(temp_dir, f'test_image_{i}.png')
    cv2.imwrite(img_path, img)
    image_paths.append(img_path)

ARGS = {
    'image_paths': image_paths
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
cv2, numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 1, dtype family float — a tuple 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_report_for_image Ran
ran on a synthesized input
ttanida/rgrg/src/full_model/generate_reports_for_images.py
code served (permissive licence) · get_code("ddfc5acae41fe5ee")
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. The code mentions cuda, and the harness aliases cuda to the CPU for every sample; whether this run needed that is 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 collections import defaultdict

torch.manual_seed(0)

# Mock tokenizer for batch_decode
class MockTokenizer:
    def batch_decode(self, token_ids, skip_special_tokens=True, clean_up_tokenization_spaces=True):
        # Return a list of decoded sentences
        return [
            "The patient has pneumonia.",
            "There is a small nodule in the right lung.",
            "The heart appears normal.",
            "Additional findings include mild consolidation."
        ]

# Mock sentence tokenizer
class MockSentenceTokenizer:
    def __call__(self, text):
        # Return an object with .sents attribute
        class SentsContainer:
            def __init__(self, text):
                self.sents = [type('Sent', (), {'text': s.strip()})() 
                             for s in text.split('.') if s.strip()]
        return SentsContainer(text)

# Mock BertScore
class MockBertScore:
    def compute(self, lang=None, predictions=None, references=None, model_type=None):
        # Return a dict with f1 scores
        return {
            'f1': [0.5]  # Below threshold, so no duplicates removed
        }

# Mock model that returns generation output
class MockImageModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(4, 8)
    
    def generate(self, image_tensor, max_length=None, num_beams=None, early_stopping=None):
        # Return a tuple: (beam_search_output, _, _, _)
        # beam_search_output should be token ids
        batch_size = image_tensor.shape[0]
        beam_search_output = torch.randint(0, 1000, (batch_size, 5))
        return (beam_search_output, None, None, None)

# Create instances
model = MockImageModel()
image_tensor = torch.randn(2, 4)  # batch_size=2, feature_dim=4
tokenizer = MockTokenizer()
bert_score = MockBertScore()
sentence_tokenizer = MockSentenceTokenizer()

ARGS = {
    'model': model,
    'image_tensor': image_tensor,
    'tokenizer': tokenizer,
    'bert_score': bert_score,
    'sentence_tokenizer': sentence_tokenizer
}
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
last_and_curr_mean_std_close Ran
ran on a synthesized input
ttanida/rgrg/src/dataset/compute_mean_std_dataset.py
code served (permissive licence) · get_code("046a23b53e85f9db")
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 bool, 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 sample data for the function parameters
curr_mean = np.array([1.0, 2.0, 3.0, 4.0])
curr_std = np.array([0.5, 0.6, 0.7, 0.8])
last_mean_values = np.array([1.00005, 2.00003, 3.00001, 4.00002])
last_std_values = np.array([0.50004, 0.60002, 0.70001, 0.80003])

ARGS = {
    "curr_mean": curr_mean,
    "curr_std": curr_std,
    "last_mean_values": last_mean_values,
    "last_std_values": last_std_values,
}
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 bool, 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
normalize_section_names Ran
ran on a synthesized input
ttanida/rgrg/src/dataset/section_parser.py
code served (permissive licence) · get_code("49696211667f2162")
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 re

def normalize_section_names(section_names):
    section_names = [s.lower().strip() for s in section_names]
    frequent_sections = {'preamble': 'preamble', 'impression': 'impression', 'comparison': 'comparison', 'indication': 'indication', 'findings': 'findings', 'examination': 'examination', 'technique': 'technique', 'history': 'history', 'comparisons': 'comparison', 'clinical history': 'history', 'reason for examination': 'indication', 'notification': 'notification', 'reason for exam': 'indication', 'clinical information': 'history', 'exam': 'examination', 'clinical indication': 'indication', 'conclusion': 'impression', 'chest, two views': 'findings', 'recommendation(s)': 'recommendations', 'type of examination': 'examination', 'reference exam': 'comparison', 'patient history': 'history', 'addendum': 'addendum', 'comparison exam': 'comparison', 'date': 'date', 'comment': 'comment', 'findings and impression': 'impression', 'wet read': 'wet read', 'comparison film': 'comparison', 'recommendations': 'recommendations', 'findings/impression': 'impression', 'pfi': 'history', 'recommendation': 'recommendations', 'wetread': 'wet read', 'ndication': 'impression', 'impresson': 'impression', 'imprression': 'impression', 'imoression': 'impression', 'impressoin': 'impression', 'imprssion': 'impression', 'impresion': 'impression', 'imperssion': 'impression', 'mpression': 'impression', 'impession': 'impression', 'findings/ impression': 'impression', 'finding': 'findings', 'findins': 'findings', 'findindgs': 'findings', 'findgings': 'findings', 'findngs': 'findings', 'findnings': 'findings', 'finidngs': 'findings', 'idication': 'indication', 'reference findings': 'findings', 'comparision': 'comparison', 'comparsion': 'comparison', 'comparrison': 'comparison', 'comparisions': 'comparison'}
    p_findings = ['chest', 'portable', 'pa and lateral', 'lateral and pa', 'ap and lateral', 'lateral and ap', 'frontal and', 'two views', 'frontal view', 'pa view', 'ap view', 'one view', 'lateral view', 'bone window', 'frontal upright', 'frontal semi-upright', 'ribs', 'pa and lat']
    p_findings = re.compile('({})'.format('|'.join(p_findings)))
    main_sections = ['impression', 'findings', 'history', 'comparison', 'addendum']
    for i, s in enumerate(section_names):
        if s in frequent_sections:
            section_names[i] = frequent_sections[s]
            continue
        main_flag = False
        for m in main_sections:
            if m in s:
                section_names[i] = m
                main_flag = True
                break
        if main_flag:
            continue
        m = p_findings.search(s)
        if m is not None:
            section_names[i] = 'findings'
    return section_names

ARGS = {
    "section_names": [
        "Impression",
        "  FINDINGS  ",
        "Clinical History",
        "PA and Lateral",
        "Unknown Section",
        "chest portable",
        "Comparisons",
        "Wet Read"
    ]
}
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
remove_duplicate_generated_sentences Ran
ran on a synthesized input
ttanida/rgrg/src/full_model/generate_reports_for_images.py
code served (permissive licence) · get_code("eeca087159c30ac6")
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
from collections import defaultdict
from unittest.mock import Mock

torch.manual_seed(0)

# Create a mock sentence tokenizer
class MockSentenceTokenizer:
    def __call__(self, text):
        # Return an object with a .sents attribute containing sentence objects
        sentences = text.split('. ')
        sent_objects = []
        for sent in sentences:
            sent_obj = Mock()
            sent_obj.text = sent.strip() + ('.' if not sent.strip().endswith('.') else '')
            sent_objects.append(sent_obj)
        
        result = Mock()
        result.sents = sent_objects
        return result

# Create a mock bert_score object
class MockBertScore:
    def compute(self, lang, predictions, references, model_type):
        # Return a dict with 'f1' key containing a list with similarity scores
        # We'll return a high score for similar sentences, low for different ones
        pred = predictions[0].lower()
        ref = references[0].lower()
        
        # Simple heuristic: if sentences are very similar or identical, high score
        if pred == ref or pred in ref or ref in pred:
            f1_score = 0.95
        else:
            f1_score = 0.5
        
        return {'f1': [f1_score]}

sentence_tokenizer = MockSentenceTokenizer()
bert_score = MockBertScore()

# Create a test generated report with some duplicate/similar sentences
generated_report = "The patient has a fever. The patient has a high temperature. The chest X-ray shows infiltrates. The lungs show infiltrates on imaging."

ARGS = {
    'generated_report': generated_report,
    'bert_score': bert_score,
    'sentence_tokenizer': sentence_tokenizer,
}
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
section_text Not yet run
unverified (harvested, no recorded run)
ttanida/rgrg/src/dataset/section_parser.py
code served (permissive licence) · get_code("93b745999721b34f")
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, 2 are also present in other papers whose code we hold (2 with at least one paper dated earlier, 0 the same month, 0 later) and 8 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.

FunctionAlso present inEarliest-datedOthers named
normalize_section_names 3 other papers
3 dated earlier
2021.acl-long.459 (2021 (year only))
Cross-modal Memory Networks for Radiology Report Generation
2107.07502 (2021-07)
MULTIBENCH: Multiscale Benchmarks for Multimodal Representation Learning
2209.07162 (2022-09)
Brain Imaging Generation with Latent Diffusion Models
section_text 2 other papers
2 dated earlier
2021.acl-long.459 (2021 (year only))
Cross-modal Memory Networks for Radiology Report Generation
2209.07162 (2022-09)
Brain Imaging Generation with Latent Diffusion Models

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

The automatic generation of radiology reports has the potential to assist radiologists in the time-consuming task of report writing. Existing methods generate the full report from image-level features, failing to explicitly focus on anatomical regions in the image. We propose a simple yet effective region-guided report generation model that detects anatomical regions and then describes individual, salient regions to form the final report. While previous methods generate reports without the possibility of human intervention and with limited explainability, our method opens up novel clinical use cases through additional interactive capabilities and introduces a high degree of transparency and explainability. Comprehensive experiments demonstrate our method's effectiveness in report generation, outperforming previous state-of-the-art models, and highlight its interactive capabilities. The code and checkpoints are available at https://github.com/ ttanida/rgrg . * Equal contribution The right lung is clear. Mild to moderate left pleural effusion as well as adjacent left basal atelectasis. Mild to moderate cardiomegaly. There is thoracolumbar dextroscoliosis. Left mid lung zone: Generated: Moderate left pleural effusion and moderate left lower lobe atelectasis are unchanged. Reference: Atelectasis of the left lower lobe with potential accompanying small left pleural effusion. Abdomen: Generated: NG tube tip is in the stomach. Reference: Nasogastric tube is in unchanged position. Cardiac silhouette: Generated: Moderate cardiomegaly persists. Reference: Unchanged moderate cardiomegaly. ground-truth prediction Right mid lung zone: Generated: Right lower lobe pneumonia. Reference: There is persistent subtle airspace opacity in the right mid to lower lung, this may reflect the residua of the patient's known pneumonia. Generated: There is no pleural effusion or pneumothorax. No acute cardiopulmonary process. Reference: The left lung is clear. No pleural effusion seen.

For agents

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

get_harvested_code_for_paper("2304.08295")
get_code_for_paper("2304.08295")
have("2304.08295")

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

curl https://syntology.ai/api/ran/2304.08295.json

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

[![Syntology run record](https://syntology.ai/api/ran/2304.08295.svg)](https://syntology.ai/paper/2304.08295)

Connect an agent — have() is free.