sample_id stringlengths 18 18 | row_index int64 0 587 | doc_id stringlengths 31 64 | page_number int64 1 12 | image_path stringlengths 34 34 | target_html stringlengths 85 12.3k | target_md stringlengths 41 7.45k | target_source stringclasses 4
values | ground_truth_method stringclasses 3
values | schema_version stringclasses 1
value | schema_version_md stringclasses 1
value | training_eligible bool 1
class | source_dataset stringclasses 1
value | arxiv_id stringclasses 60
values | components_json stringclasses 56
values | styles_json stringclasses 89
values | content_brief_json stringclasses 60
values | llm_providers_json stringclasses 30
values |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
synthetic_00000000 | 0 | full240_2026-10-14/sample_00002 | 1 | page-images/synthetic_00000000.png | <div data-bbox="[218, 204, 782, 223]" data-label="Section-Header"><h1>ChunkAct: Action Chunking for Agentic Robotics</h1></div>
<div data-bbox="[444, 277, 555, 291]" data-label="Text"><p>23 April 2023</p></div>
<div data-bbox="[120, 328, 413, 345]" data-label="Section-Header"><h2>Typed Exit Taxonomy</h2></div>
<div dat... | 
# ChunkAct: Action Chunking for Agentic Robotics
23 April 2023
## Typed Exit Taxonomy
1 lists all exit codes that a chunk may return upon failure, with its meaning and expected planner reaction.
![heading](119,... | synthetic_layout_real_text | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18874 | [] | {"font_size": 11, "font_format": "serif", "table_style": "striped", "figure_style": "borderless", "text_color": "dark_blue", "column_layout": "two_column"} | {} | [] |
synthetic_00000001 | 1 | full240_2026-10-14/sample_00003 | 1 | page-images/synthetic_00000001.png | <div data-bbox="[230, 219, 768, 305]" data-label="Section-Header"><h1>Risk Assessment of 19th Century Opium Trade Losses in Colonial Singapore Port 1825</h1></div>
<div data-bbox="[397, 366, 600, 381]" data-label="Text"><p>September 1, 2024</p></div>
<div data-bbox="[184, 430, 372, 449]" data-label="Section-Header"><h2... | 
# Risk Assessment of 19th Century Opium Trade Losses in Colonial Singapore Port 1825
September 1, 2024
## Background
The opium consignment of 1825 illustrated how cargo valuation depended on sequential handling ... | synthetic_both | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18874 | ["bibliography", "code_block", "figure", "formula", "page_header"] | {"font_size": 12, "font_format": "monospace", "table_style": "striped", "figure_style": "borderless", "text_color": "dark_gray", "column_layout": "one_column"} | {"genre": "security advisory", "register": "practical and instructional, addressing the reader directly", "angle": "a historical account of how the practice developed", "setting": "an insurance firm in Singapore", "title_style": "a long specific title of 9 to 14 words naming the setting or year", "section_count": 3, "p... | [] |
synthetic_00000002 | 2 | full240_2026-10-14/sample_00003 | 2 | page-images/synthetic_00000002.png | <div data-bbox="[182, 160, 606, 179]" data-label="Section-Header"><h2>Mitigation Recommendations</h2></div>
<div data-bbox="[181, 198, 801, 265]" data-label="Text"><p>We advise clients to embed real-time ledger updates and to allocate a contingency reserve of at least five percent of cargo value, computed via the loss ... | 
## Mitigation Recommendations
We advise clients to embed real-time ledger updates and to allocate a contingency reserve of at least five percent of cargo value, computed via the loss function L = P * (1 - r) where P is the declared price and r the historical loss ra... | synthetic_both | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18874 | ["bibliography", "code_block", "figure", "formula", "page_header"] | {"font_size": 12, "font_format": "monospace", "table_style": "striped", "figure_style": "borderless", "text_color": "dark_gray", "column_layout": "one_column"} | {"genre": "security advisory", "register": "practical and instructional, addressing the reader directly", "angle": "a historical account of how the practice developed", "setting": "an insurance firm in Singapore", "title_style": "a long specific title of 9 to 14 words naming the setting or year", "section_count": 3, "p... | [] |
synthetic_00000003 | 3 | full240_2026-10-14/sample_00006 | 1 | page-images/synthetic_00000003.png | <div data-bbox="[119, 202, 880, 257]" data-label="Section-Header"><h1>LoopWAM: Closed-Loop Generalist Policy Improvement from Deployment Rollouts</h1></div>
<div data-bbox="[429, 320, 569, 337]" data-label="Text"><p>April 28, 2016</p></div>
<div data-bbox="[120, 376, 490, 419]" data-label="Section-Header"><h2>Failure R... | 
# LoopWAM: Closed-Loop Generalist Policy Improvement from Deployment Rollouts
April 28, 2016
## Failure Recovery and Iterative Execution
Real-world robot manipulation often involves uncertainty caused by imperfe... | synthetic_layout_real_text | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18873 | [] | {"font_size": 12, "font_format": "serif", "table_style": "striped", "figure_style": "borderless", "text_color": "black", "column_layout": "two_column"} | {} | [] |
synthetic_00000004 | 4 | full240_2026-10-14/sample_00007 | 1 | page-images/synthetic_00000004.png | <div data-bbox="[450, 209, 549, 228]" data-label="Section-Header"><h1>Signage</h1></div>
<div data-bbox="[445, 278, 554, 291]" data-label="Text"><p>23 May 2025</p></div>
<div data-bbox="[221, 330, 339, 345]" data-label="Section-Header"><h2>Purpose</h2></div>
<div data-bbox="[219, 361, 773, 626]" data-label="Text"><p>Th... | 
# Signage
23 May 2025
## Purpose
The objective of this memorandum is to outline the step‑by‑step protocol for approving new public signage typography for the Gauteng Provincial Court premises, with specific refe... | synthetic_both | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18873 | ["formula", "page_header", "table"] | {"font_size": 10, "font_format": "monospace", "table_style": "borderless", "figure_style": "borderless", "text_color": "black", "column_layout": "one_column"} | {"genre": "internal memo", "register": "terse bulletin style, fragments allowed where natural", "angle": "a step-by-step procedure", "setting": "a legal practice in South Africa", "title_style": "a short label of 1 to 3 words", "section_count": 3, "paragraphs_per_section": 1, "sentences_per_paragraph": 5, "textures": [... | [] |
synthetic_00000005 | 5 | full240_2026-10-14/sample_00007 | 2 | page-images/synthetic_00000005.png | <div data-bbox="[219, 162, 774, 291]" data-label="Text"><p>language is accurately reproduced. Step four: the approved file is sent to the contracted sign fabricator, SignTech Ltd., with the purchase order referencing product code ST‑SIGN‑HV2026, which denotes the high-visibility aluminium composite material required fo... | 
language is accurately reproduced. Step four: the approved file is sent to the contracted sign fabricator, SignTech Ltd., with the purchase order referencing product code ST‑SIGN‑HV2026, which denotes the high-visibility aluminium composite material required for durability. Step five: upon rece... | synthetic_both | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18873 | ["formula", "page_header", "table"] | {"font_size": 10, "font_format": "monospace", "table_style": "borderless", "figure_style": "borderless", "text_color": "black", "column_layout": "one_column"} | {"genre": "internal memo", "register": "terse bulletin style, fragments allowed where natural", "angle": "a step-by-step procedure", "setting": "a legal practice in South Africa", "title_style": "a short label of 1 to 3 words", "section_count": 3, "paragraphs_per_section": 1, "sentences_per_paragraph": 5, "textures": [... | [] |
synthetic_00000006 | 6 | full240_2026-10-14/sample_00010 | 1 | page-images/synthetic_00000006.png | <div data-bbox="[239, 207, 760, 279]" data-label="Section-Header"><h1>EgoExplorer: A Vision-Motion-Touch Observatory for Human Loco-Manipulation In Situ</h1></div>
<div data-bbox="[433, 334, 567, 348]" data-label="Text"><p>16 January 2009</p></div>
<div data-bbox="[221, 386, 397, 399]" data-label="Section-Header"><h2>T... | 
# EgoExplorer: A Vision-Motion-Touch Observatory for Human Loco-Manipulation In Situ
16 January 2009
## The Dataset
Egocentric video captures task semantics at scale, yet leaves key signals for under-constrained... | synthetic_layout_real_text | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18872 | [] | {"font_size": 10, "font_format": "serif", "table_style": "striped", "figure_style": "borderless", "text_color": "dark_gray", "column_layout": "one_column"} | {} | [] |
synthetic_00000007 | 7 | full240_2026-10-14/sample_00010 | 2 | page-images/synthetic_00000007.png | <div data-bbox="[220, 158, 400, 170]" data-label="Section-Header"><h2>Introduction</h2></div>
<div data-bbox="[218, 189, 780, 336]" data-label="Text"><p>What a humanoid needs to reach, grasp, or carry shapes where it moves and how it positions its body [1]. Learning such skills calls for demonstrations spanning diverse... | 
## Introduction
What a humanoid needs to reach, grasp, or carry shapes where it moves and how it positions its body [1]. Learning such skills calls for demonstrations spanning diverse environments, objects, and task sequences. Robot teleoperation provides such data,... | synthetic_layout_real_text | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18872 | [] | {"font_size": 10, "font_format": "serif", "table_style": "striped", "figure_style": "borderless", "text_color": "dark_gray", "column_layout": "one_column"} | {} | [] |
synthetic_00000008 | 8 | full240_2026-10-14/sample_00011 | 1 | page-images/synthetic_00000008.png | <div data-bbox="[225, 209, 775, 251]" data-label="Section-Header"><h1>2023 Deployment of Adaptive Congestion Control for Metrobus Data Links</h1></div>
<div data-bbox="[420, 305, 579, 318]" data-label="Text"><p>23 February 2013</p></div>
<div data-bbox="[221, 358, 536, 373]" data-label="Section-Header"><h2>Background a... | 
# 2023 Deployment of Adaptive Congestion Control for Metrobus Data Links
23 February 2013
## Background and Incident
In March 2023 the Sistema de Transporte Colectivo (STC) in Mexico City reported severe packet ... | synthetic_both | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18872 | ["bibliography", "code_block", "figure", "formula", "page_header", "table"] | {"font_size": 10, "font_format": "monospace", "table_style": "borderless", "figure_style": "bordered", "text_color": "dark_gray", "column_layout": "one_column"} | {"genre": "conference abstract", "register": "narrative, recounting events in order", "angle": "a single case described in detail", "setting": "a public transit authority in Mexico", "title_style": "a title starting with a number or a year", "section_count": 2, "paragraphs_per_section": 1, "sentences_per_paragraph": 5,... | [] |
synthetic_00000009 | 9 | full240_2026-10-14/sample_00011 | 2 | page-images/synthetic_00000009.png | <div data-bbox="[219, 162, 774, 233]" data-label="Text"><p>consistent across all monitored routes, and the transit authority reported a 15 percent increase in on‑time arrivals for the following week. These results demonstrate that targeted congestion control can stabilize urban transit communication networks without ha... | 
consistent across all monitored routes, and the transit authority reported a 15 percent increase in on‑time arrivals for the following week. These results demonstrate that targeted congestion control can stabilize urban transit communication networks without hardware upgrades.
![table](228,251... | synthetic_both | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18872 | ["bibliography", "code_block", "figure", "formula", "page_header", "table"] | {"font_size": 10, "font_format": "monospace", "table_style": "borderless", "figure_style": "bordered", "text_color": "dark_gray", "column_layout": "one_column"} | {"genre": "conference abstract", "register": "narrative, recounting events in order", "angle": "a single case described in detail", "setting": "a public transit authority in Mexico", "title_style": "a title starting with a number or a year", "section_count": 2, "paragraphs_per_section": 1, "sentences_per_paragraph": 5,... | [] |
synthetic_00000010 | 10 | full240_2026-10-14/sample_00014 | 1 | page-images/synthetic_00000010.png | <div data-bbox="[219, 217, 778, 265]" data-label="Section-Header"><h1>Dog-Breed Text-to-image Flow Matching from Pretraining to Fine-Tuning on 4 Consumer GPUs</h1></div>
<div data-bbox="[435, 319, 562, 333]" data-label="Text"><p>March 31, 2009</p></div>
<div data-bbox="[208, 376, 326, 389]" data-label="Section-Header">... | 
# Dog-Breed Text-to-image Flow Matching from Pretraining to Fine-Tuning on 4 Consumer GPUs
March 31, 2009
## Results
Quality of generated images. Our best pretraining recipe is to use E2E-INVAE as tokenizer, 128... | synthetic_layout_real_text | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18870 | [] | {"font_size": 11, "font_format": "serif", "table_style": "striped", "figure_style": "bordered", "text_color": "dark_blue", "column_layout": "one_column"} | {} | [] |
synthetic_00000011 | 11 | full240_2026-10-14/sample_00015 | 1 | page-images/synthetic_00000011.png | <div data-bbox="[214, 217, 783, 265]" data-label="Section-Header"><h1>Field Study of Circadian Gene GENE1 Expression in Rural Indian Cooperative, 2023</h1></div>
<div data-bbox="[429, 319, 567, 330]" data-label="Text"><p>5 December 2011</p></div>
<div data-bbox="[207, 372, 368, 389]" data-label="Section-Header"><h2>Bac... | 
# Field Study of Circadian Gene GENE1 Expression in Rural Indian Cooperative, 2023
5 December 2011
## Background
In early 2023 the Madhura Cooperative of Uttar Pradesh agreed to host a monitoring project on nigh... | synthetic_both | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18870 | ["bibliography", "code_block", "formula", "list", "page_header", "table"] | {"font_size": 11, "font_format": "sans_serif", "table_style": "striped", "figure_style": "bordered", "text_color": "dark_gray", "column_layout": "one_column"} | {"genre": "museum exhibit catalogue entry", "register": "narrative, recounting events in order", "angle": "a measurement study with concrete results", "setting": "a rural cooperative in India", "title_style": "a long specific title of 9 to 14 words naming the setting or year", "section_count": 7, "paragraphs_per_sectio... | [] |
synthetic_00000012 | 12 | full240_2026-10-14/sample_00015 | 2 | page-images/synthetic_00000012.png | <div data-bbox="[206, 165, 439, 182]" data-label="Section-Header"><h2>Sampling Protocol</h2></div>
<div data-bbox="[206, 199, 790, 229]" data-label="Text"><p>Researchers visited each household twice a week, collecting peripheral blood samples at 22:00 and again at 02:00 during the monsoon season.</p></div>
<div data-bb... | 
## Sampling Protocol
Researchers visited each household twice a week, collecting peripheral blood samples at 22:00 and again at 02:00 during the monsoon season.
All samples were stored in portable coolers at four degrees Celsius and later a... | synthetic_both | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18870 | ["bibliography", "code_block", "formula", "list", "page_header", "table"] | {"font_size": 11, "font_format": "sans_serif", "table_style": "striped", "figure_style": "bordered", "text_color": "dark_gray", "column_layout": "one_column"} | {"genre": "museum exhibit catalogue entry", "register": "narrative, recounting events in order", "angle": "a measurement study with concrete results", "setting": "a rural cooperative in India", "title_style": "a long specific title of 9 to 14 words naming the setting or year", "section_count": 7, "paragraphs_per_sectio... | [] |
synthetic_00000013 | 13 | full240_2026-10-14/sample_00015 | 3 | page-images/synthetic_00000013.png | <div data-bbox="[206, 165, 447, 182]" data-label="Section-Header"><h2>Community Impact</h2></div>
<div data-bbox="[206, 199, 791, 229]" data-label="Text"><p>Following the report, the cooperative elected to adjust evening lighting, reducing artificial illumination after 22:00 to preserve natural melatonin cues.</p></div... | 
## Community Impact
Following the report, the cooperative elected to adjust evening lighting, reducing artificial illumination after 22:00 to preserve natural melatonin cues.
Health workshops incorporated the message that consistent sleep d... | synthetic_both | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18870 | ["bibliography", "code_block", "formula", "list", "page_header", "table"] | {"font_size": 11, "font_format": "sans_serif", "table_style": "striped", "figure_style": "bordered", "text_color": "dark_gray", "column_layout": "one_column"} | {"genre": "museum exhibit catalogue entry", "register": "narrative, recounting events in order", "angle": "a measurement study with concrete results", "setting": "a rural cooperative in India", "title_style": "a long specific title of 9 to 14 words naming the setting or year", "section_count": 7, "paragraphs_per_sectio... | [] |
synthetic_00000014 | 14 | full240_2026-10-14/sample_00018 | 1 | page-images/synthetic_00000014.png | <div data-bbox="[120, 195, 879, 242]" data-label="Section-Header"><h1>Cepheid: Decoupling Scale from Structure for Cross-Family Motion Completion</h1></div>
<div data-bbox="[433, 294, 567, 307]" data-label="Text"><p>7 February 2023</p></div>
<div data-bbox="[120, 339, 244, 352]" data-label="Section-Header"><h2>Method</... | 
# Cepheid: Decoupling Scale from Structure for Cross-Family Motion Completion
7 February 2023
## Method
Let M = (V, E) be a surface mesh with N = |V| vertices and edge set E, and let X RN 3 denote its source-fra... | synthetic_layout_real_text | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18871 | ["table"] | {"font_size": 10, "font_format": "serif", "table_style": "borderless", "figure_style": "borderless", "text_color": "black", "column_layout": "two_column"} | {} | [] |
synthetic_00000015 | 15 | full240_2026-10-14/sample_00018 | 2 | page-images/synthetic_00000015.png | <div data-bbox="[118, 161, 223, 170]" data-label="Text"><p>across families.</p></div>
<div data-bbox="[118, 199, 239, 212]" data-label="Section-Header"><h2>Results</h2></div>
<div data-bbox="[118, 230, 491, 348]" data-label="Text"><p>Our model achieves lower error than NSDP and DOMA at all five densities on DT4D, landi... | 
across families.
## Results
Our model achieves lower error than NSDP and DOMA at all five densities on DT4D, landing 44--81% below harmonic extension (Table 1). All three methods are evaluated on the same seven held-out identities of one sp... | synthetic_layout_real_text | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18871 | ["table"] | {"font_size": 10, "font_format": "serif", "table_style": "borderless", "figure_style": "borderless", "text_color": "black", "column_layout": "two_column"} | {} | [] |
synthetic_00000016 | 16 | full240_2026-10-14/sample_00019 | 1 | page-images/synthetic_00000016.png | <div data-bbox="[144, 204, 855, 251]" data-label="Section-Header"><h1>Assessing Referee Decision Accuracy with Video Review: A Case Study of the 2023 J-League Penalty Review System</h1></div>
<div data-bbox="[446, 305, 552, 319]" data-label="Text"><p>July 19, 2000</p></div>
<div data-bbox="[119, 356, 433, 372]" data-la... | 
# Assessing Referee Decision Accuracy with Video Review: A Case Study of the 2023 J-League Penalty Review System
July 19, 2000
## Results and Interpretation
The analysis of the 2023 J-League penalty review syste... | synthetic_both | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18871 | ["bibliography", "code_block", "list", "page_footer", "page_header"] | {"font_size": 11, "font_format": "sans_serif", "table_style": "striped", "figure_style": "bordered", "text_color": "black", "column_layout": "two_column"} | {"genre": "thesis chapter", "register": "formal, for specialists, dense with terminology", "angle": "a single case described in detail", "setting": "a national statistics agency in Japan", "title_style": "a title with a colon and a subtitle", "section_count": 1, "paragraphs_per_section": 3, "sentences_per_paragraph": 4... | [] |
synthetic_00000017 | 17 | full240_2026-10-14/sample_00022 | 1 | page-images/synthetic_00000017.png | <div data-bbox="[188, 217, 812, 272]" data-label="Section-Header"><h1>What Matters in Quality-Aware VLA Pretraining: A Benchmark and Empirical Study</h1></div>
<div data-bbox="[414, 336, 585, 352]" data-label="Text"><p>February 12, 2014</p></div>
<div data-bbox="[183, 400, 529, 420]" data-label="Section-Header"><h2>Sim... | 
# What Matters in Quality-Aware VLA Pretraining: A Benchmark and Empirical Study
February 12, 2014
## Simulation Experiments
We ask three questions: which quality-aware pretraining choices improve performance ac... | synthetic_layout_real_text | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18869 | [] | {"font_size": 12, "font_format": "sans_serif", "table_style": "borderless", "figure_style": "bordered", "text_color": "dark_blue", "column_layout": "one_column"} | {} | [] |
synthetic_00000018 | 18 | full240_2026-10-14/sample_00022 | 2 | page-images/synthetic_00000018.png | <div data-bbox="[182, 158, 382, 175]" data-label="Section-Header"><h2>Introduction</h2></div>
<div data-bbox="[181, 197, 816, 321]" data-label="Text"><p>Vision-language-action (VLA) policies are commonly pretrained on multitask demonstrations and fine-tuned on target tasks before deployment [1, 2, 3, 4]. During a demon... | 
## Introduction
Vision-language-action (VLA) policies are commonly pretrained on multitask demonstrations and fine-tuned on target tasks before deployment [1, 2, 3, 4]. During a demonstration, an operator may hesitate or move an object along an unnecessary detour, p... | synthetic_layout_real_text | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18869 | [] | {"font_size": 12, "font_format": "sans_serif", "table_style": "borderless", "figure_style": "bordered", "text_color": "dark_blue", "column_layout": "one_column"} | {} | [] |
synthetic_00000019 | 19 | full240_2026-10-14/sample_00027 | 1 | page-images/synthetic_00000019.png | <div data-bbox="[364, 217, 635, 236]" data-label="Section-Header"><h1>Remote Work Terms</h1></div>
<div data-bbox="[446, 304, 551, 317]" data-label="Text"><p>2005-03-06</p></div>
<div data-bbox="[183, 358, 627, 378]" data-label="Section-Header"><h2>Definition of Remote Employee</h2></div>
<div data-bbox="[181, 397, 816... | 
# Remote Work Terms
2005-03-06
## Definition of Remote Employee
A Remote Employee is a staff member who performs contractual duties outside the principal place of business, typically from a personal residence or... | synthetic_both | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18868 | ["footnote", "list"] | {"font_size": 12, "font_format": "sans_serif", "table_style": "ruled", "figure_style": "bordered", "text_color": "dark_gray", "column_layout": "one_column"} | {"genre": "legal contract clause commentary", "register": "journalistic, with concrete examples and quotes", "angle": "a glossary-like explanation of key terms", "setting": "a logistics company in Kenya", "title_style": "a short label of 1 to 3 words", "section_count": 7, "paragraphs_per_section": 1, "sentences_per_par... | [] |
synthetic_00000020 | 20 | full240_2026-10-14/sample_00027 | 2 | page-images/synthetic_00000020.png | <div data-bbox="[182, 159, 488, 179]" data-label="Section-Header"><h2>Work‑Hour Tracking</h2></div>
<div data-bbox="[181, 197, 817, 284]" data-label="Text"><p>All remote staff are required to log hours through the approved time‑keeping portal, with a minimum of eight hours per day and a maximum of ten hours, a range de... | 
## Work‑Hour Tracking
All remote staff are required to log hours through the approved time‑keeping portal, with a minimum of eight hours per day and a maximum of ten hours, a range designed to avoid overtime abuse. The system automatically flags any day where logged... | synthetic_both | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18868 | ["footnote", "list"] | {"font_size": 12, "font_format": "sans_serif", "table_style": "ruled", "figure_style": "bordered", "text_color": "dark_gray", "column_layout": "one_column"} | {"genre": "legal contract clause commentary", "register": "journalistic, with concrete examples and quotes", "angle": "a glossary-like explanation of key terms", "setting": "a logistics company in Kenya", "title_style": "a short label of 1 to 3 words", "section_count": 7, "paragraphs_per_section": 1, "sentences_per_par... | [] |
synthetic_00000021 | 21 | full240_2026-10-14/sample_00030 | 1 | page-images/synthetic_00000021.png | <div data-bbox="[276, 209, 724, 283]" data-label="Section-Header"><h1>PLAYGROUND: Progressive Layout Generation with Render-Grounded Planning</h1></div>
<div data-bbox="[450, 333, 549, 343]" data-label="Text"><p>2004-09-24</p></div>
<div data-bbox="[221, 385, 537, 397]" data-label="Section-Header"><h2>Additional Relate... | 
# PLAYGROUND: Progressive Layout Generation with Render-Grounded Planning
2004-09-24
## Additional Related Work
This section provides additional discussion of MLLMs and broader graphic design layout generation b... | synthetic_layout_real_text | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18865 | [] | {"font_size": 10, "font_format": "monospace", "table_style": "striped", "figure_style": "borderless", "text_color": "black", "column_layout": "one_column"} | {} | [] |
synthetic_00000022 | 22 | full240_2026-10-14/sample_00030 | 2 | page-images/synthetic_00000022.png | <div data-bbox="[219, 162, 730, 188]" data-label="Text"><p>input assumptions of each method while using the same source design content.</p></div>
<div data-bbox="[219, 192, 773, 294]" data-label="Text"><p>LICA [18] does not provide document titles, which are required as metadata inputs to our pipeline. We therefore gen... | 
input assumptions of each method while using the same source design content.
LICA [18] does not provide document titles, which are required as metadata inputs to our pipeline. We therefore generate a title for each LICA sample using Qwen3.5-27B [33]. We use the ground-... | synthetic_layout_real_text | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18865 | [] | {"font_size": 10, "font_format": "monospace", "table_style": "striped", "figure_style": "borderless", "text_color": "black", "column_layout": "one_column"} | {} | [] |
synthetic_00000023 | 23 | full240_2026-10-14/sample_00031 | 1 | page-images/synthetic_00000023.png | <div data-bbox="[370, 204, 627, 226]" data-label="Section-Header"><h1>Housing Options</h1></div>
<div data-bbox="[415, 288, 583, 302]" data-label="Text"><p>March 12, 2006</p></div>
<div data-bbox="[121, 342, 278, 356]" data-label="Section-Header"><h2>Overview</h2></div>
<div data-bbox="[118, 379, 490, 465]" data-label=... | 
# Housing Options
March 12, 2006
## Overview
The Regional Hospital in Da Nang is evaluating two redevelopment schemes for the old textile mill site, aiming to provide affordable staff housing while preserving gr... | synthetic_both | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18865 | ["bibliography", "code_block", "formula", "list"] | {"font_size": 12, "font_format": "monospace", "table_style": "ruled", "figure_style": "bordered", "text_color": "dark_gray", "column_layout": "two_column"} | {"genre": "customer support knowledge base article", "register": "journalistic, with concrete examples and quotes", "angle": "a comparison of two or three alternatives", "setting": "a regional hospital in Vietnam", "title_style": "a short label of 1 to 3 words", "section_count": 3, "paragraphs_per_section": 2, "sentenc... | [] |
synthetic_00000024 | 24 | full240_2026-10-14/sample_00031 | 2 | page-images/synthetic_00000024.png | <div data-bbox="[118, 164, 419, 193]" data-label="Text"><p>favor Option A for its quieter environment.</p></div>
<div data-bbox="[118, 229, 263, 243]" data-label="Section-Header"><h2>References</h2></div>
<div data-bbox="[122, 265, 489, 295]" data-label="Text"><p>[1] Le et al. Redeveloping Industrial Land in Vietnam. 2... | 
favor Option A for its quieter environment.
## References
[1] Le et al. Redeveloping Industrial Land in Vietnam. 2022.
[2] Pham and Ho. Cost Analysis of Affordable Housing. 2021.
[3] World... | synthetic_both | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18865 | ["bibliography", "code_block", "formula", "list"] | {"font_size": 12, "font_format": "monospace", "table_style": "ruled", "figure_style": "bordered", "text_color": "dark_gray", "column_layout": "two_column"} | {"genre": "customer support knowledge base article", "register": "journalistic, with concrete examples and quotes", "angle": "a comparison of two or three alternatives", "setting": "a regional hospital in Vietnam", "title_style": "a short label of 1 to 3 words", "section_count": 3, "paragraphs_per_section": 2, "sentenc... | [] |
synthetic_00000025 | 25 | full240_2026-10-14/sample_00034 | 1 | page-images/synthetic_00000025.png | <div data-bbox="[225, 209, 773, 251]" data-label="Section-Header"><h1>Data Free Training of Diffusion Models for Lattice Field Theories</h1></div>
<div data-bbox="[431, 305, 569, 318]" data-label="Text"><p>1 January 2022</p></div>
<div data-bbox="[221, 358, 697, 396]" data-label="Section-Header"><h2>Lattice sizes, phas... | 
# Data Free Training of Diffusion Models for Lattice Field Theories
1 January 2022
## Lattice sizes, phases, network architecture, and training algorithm
All experiments use the D=2 theory on periodic square lat... | synthetic_layout_real_text | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18863 | ["table"] | {"font_size": 10, "font_format": "monospace", "table_style": "striped", "figure_style": "borderless", "text_color": "dark_gray", "column_layout": "one_column"} | {} | [] |
synthetic_00000026 | 26 | full240_2026-10-14/sample_00034 | 2 | page-images/synthetic_00000026.png | <div data-bbox="[243, 159, 757, 216]" data-label="Table"><table><thead><tr><th>L</th><th>phase</th><th>F/||</th><th>fTI</th><th>ESS (%)</th><th>z</th></tr></thead><tbody><tr><td>4</td><td>symmetric</td><td>-0.687128(13)</td><td>-0.687127(39)</td><td>99.8</td><td>-0.03</td></tr><tr><td>8</td><td>symmetric</td><td>-0.670... | 
<table><tr><th>L</th><th>phase</th><th>F/||</th><th>fTI</th><th>ESS (%)</th><th>z</th></tr><tr><td>4</td><td>symmetric</td><td>-0.687128(13)</td><td>-0.687127(39)</td><td>99.8</td><td>-0.03</td></tr><tr><td>8</td><td>symmetric</td><td>-0.670360(7)</td><td>-0.670380(17)</td><td>99.0</td><td>+1.... | synthetic_layout_real_text | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18863 | ["table"] | {"font_size": 10, "font_format": "monospace", "table_style": "striped", "figure_style": "borderless", "text_color": "dark_gray", "column_layout": "one_column"} | {} | [] |
synthetic_00000027 | 27 | full240_2026-10-14/sample_00035 | 1 | page-images/synthetic_00000027.png | <div data-bbox="[241, 217, 758, 289]" data-label="Section-Header"><h1>Is the School Lunch Nutrition Policy Improving Student Health at the Westfield Mine Primary School?</h1></div>
<div data-bbox="[446, 347, 551, 358]" data-label="Text"><p>15 June 2003</p></div>
<div data-bbox="[207, 404, 453, 420]" data-label="Section... | 
# Is the School Lunch Nutrition Policy Improving Student Health at the Westfield Mine Primary School?
15 June 2003
## Executive Summary
The pilot nutrition policy was launched on 1 March 2023 at the Westfield Mi... | synthetic_both | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18863 | ["bibliography", "footnote", "formula", "page_footer"] | {"font_size": 11, "font_format": "sans_serif", "table_style": "borderless", "figure_style": "borderless", "text_color": "black", "column_layout": "one_column"} | {"genre": "project status report", "register": "plain and administrative, short declarative sentences", "angle": "a review of a policy and its effects", "setting": "a mining operation in Australia", "title_style": "a title phrased as a question", "section_count": 6, "paragraphs_per_section": 3, "sentences_per_paragraph... | [] |
synthetic_00000028 | 28 | full240_2026-10-14/sample_00035 | 2 | page-images/synthetic_00000028.png | <div data-bbox="[206, 165, 382, 182]" data-label="Section-Header"><h2>Methodology</h2></div>
<div data-bbox="[206, 199, 791, 247]" data-label="Text"><p>Data were collected monthly from school meal logs, with each entry capturing portion size, nutrient content, and student consumption rates, using the standardized Nutri... | 
## Methodology
Data were collected monthly from school meal logs, with each entry capturing portion size, nutrient content, and student consumption rates, using the standardized Nutrition Tracking System (NTS) version 3.2.
A control group o... | synthetic_both | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18863 | ["bibliography", "footnote", "formula", "page_footer"] | {"font_size": 11, "font_format": "sans_serif", "table_style": "borderless", "figure_style": "borderless", "text_color": "black", "column_layout": "one_column"} | {"genre": "project status report", "register": "plain and administrative, short declarative sentences", "angle": "a review of a policy and its effects", "setting": "a mining operation in Australia", "title_style": "a title phrased as a question", "section_count": 6, "paragraphs_per_section": 3, "sentences_per_paragraph... | [] |
synthetic_00000029 | 29 | full240_2026-10-14/sample_00035 | 3 | page-images/synthetic_00000029.png | <div data-bbox="[206, 165, 437, 178]" data-label="Section-Header"><h2>Recommendations</h2></div>
<div data-bbox="[205, 199, 791, 246]" data-label="Text"><p>We recommend extending the nutrition education module to include a hands‑on cooking workshop each semester, which research indicates can improve willingness to try ... | 
## Recommendations
We recommend extending the nutrition education module to include a hands‑on cooking workshop each semester, which research indicates can improve willingness to try new foods by up to 25 percent.
The school should implemen... | synthetic_both | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18863 | ["bibliography", "footnote", "formula", "page_footer"] | {"font_size": 11, "font_format": "sans_serif", "table_style": "borderless", "figure_style": "borderless", "text_color": "black", "column_layout": "one_column"} | {"genre": "project status report", "register": "plain and administrative, short declarative sentences", "angle": "a review of a policy and its effects", "setting": "a mining operation in Australia", "title_style": "a title phrased as a question", "section_count": 6, "paragraphs_per_section": 3, "sentences_per_paragraph... | [] |
synthetic_00000030 | 30 | full240_2026-10-14/sample_00038 | 1 | page-images/synthetic_00000030.png | <div data-bbox="[236, 219, 760, 266]" data-label="Section-Header"><h1>DoubtLess: Training-Free Contrastive Decoding for Efficient Reasoning</h1></div>
<div data-bbox="[449, 320, 548, 330]" data-label="Text"><p>2003-07-23</p></div>
<div data-bbox="[208, 377, 658, 415]" data-label="Section-Header"><h2>The logit differenc... | 
# DoubtLess: Training-Free Contrastive Decoding for Efficient Reasoning
2003-07-23
## The logit difference as a ratio of probabilities
Write pM, pb and pv for the softmax distributions of , and , and ZM, Zb, Zv ... | synthetic_layout_real_text | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18862 | ["table"] | {"font_size": 11, "font_format": "monospace", "table_style": "ruled", "figure_style": "borderless", "text_color": "black", "column_layout": "one_column"} | {} | [] |
synthetic_00000031 | 31 | full240_2026-10-14/sample_00038 | 2 | page-images/synthetic_00000031.png | <div data-bbox="[207, 166, 389, 178]" data-label="Section-Header"><h2>Related work</h2></div>
<div data-bbox="[206, 200, 786, 294]" data-label="Text"><p>Reasoning models spend a large share of their tokens re-deriving and re-verifying results they have already reached, even on simple problems [1, 2], and their token sp... | 
## Related work
Reasoning models spend a large share of their tokens re-deriving and re-verifying results they have already reached, even on simple problems [1, 2], and their token spend tracks the difficulty of a problem only loosely [33]. Section 1 shows that the ... | synthetic_layout_real_text | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18862 | ["table"] | {"font_size": 11, "font_format": "monospace", "table_style": "ruled", "figure_style": "borderless", "text_color": "black", "column_layout": "one_column"} | {} | [] |
synthetic_00000032 | 32 | full240_2026-10-14/sample_00038 | 3 | page-images/synthetic_00000032.png | <div data-bbox="[206, 169, 776, 198]" data-label="Text"><p>strength at decoding time and can be added to a model trained this way (Section 1).</p></div> | 
strength at decoding time and can be added to a model trained this way (Section 1). | synthetic_layout_real_text | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18862 | ["table"] | {"font_size": 11, "font_format": "monospace", "table_style": "ruled", "figure_style": "borderless", "text_color": "black", "column_layout": "one_column"} | {} | [] |
synthetic_00000033 | 33 | full240_2026-10-14/sample_00039 | 1 | page-images/synthetic_00000033.png | <div data-bbox="[140, 202, 859, 257]" data-label="Section-Header"><h1>Threading Performance: A Textile Exporter's Guide to Index Tuning for Write-Heavy Order Systems</h1></div>
<div data-bbox="[433, 320, 567, 336]" data-label="Text"><p>June 29, 2024</p></div>
<div data-bbox="[120, 374, 414, 394]" data-label="Section-He... | 
# Threading Performance: A Textile Exporter's Guide to Index Tuning for Write-Heavy Order Systems
June 29, 2024
## Executive Summary
Bangladesh's leading garment exporter, Shanta Textiles, recently overhauled it... | synthetic_both | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18862 | ["figure"] | {"font_size": 12, "font_format": "sans_serif", "table_style": "striped", "figure_style": "borderless", "text_color": "dark_blue", "column_layout": "two_column"} | {"genre": "cooking and nutrition guide", "register": "journalistic, with concrete examples and quotes", "angle": "a review of a policy and its effects", "setting": "a textile exporter in Bangladesh", "title_style": "a title with a colon and a subtitle", "section_count": 9, "paragraphs_per_section": 2, "sentences_per_pa... | [] |
synthetic_00000034 | 34 | full240_2026-10-14/sample_00039 | 2 | page-images/synthetic_00000034.png | <div data-bbox="[118, 163, 490, 195]" data-label="Text"><p>hour without a single timeout, a 10 to 15 percent improvement over the previous peak.</p></div>
<div data-bbox="[118, 201, 491, 267]" data-label="Text"><p>Production manager Laila Hossain noted that order confirmations now reach customers within two minutes, co... | 
hour without a single timeout, a 10 to 15 percent improvement over the previous peak.
Production manager Laila Hossain noted that order confirmations now reach customers within two minutes, compared with the earlier five‑minute window.
## C... | synthetic_both | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18862 | ["figure"] | {"font_size": 12, "font_format": "sans_serif", "table_style": "striped", "figure_style": "borderless", "text_color": "dark_blue", "column_layout": "two_column"} | {"genre": "cooking and nutrition guide", "register": "journalistic, with concrete examples and quotes", "angle": "a review of a policy and its effects", "setting": "a textile exporter in Bangladesh", "title_style": "a title with a colon and a subtitle", "section_count": 9, "paragraphs_per_section": 2, "sentences_per_pa... | [] |
synthetic_00000035 | 35 | full240_2026-10-14/sample_00042 | 1 | page-images/synthetic_00000035.png | <div data-bbox="[335, 217, 665, 241]" data-label="Section-Header"><h1>Hybrid Reasoning Agents</h1></div>
<div data-bbox="[422, 304, 576, 320]" data-label="Text"><p>4 February 2024</p></div>
<div data-bbox="[183, 368, 454, 389]" data-label="Section-Header"><h2>In-depth Analyses</h2></div>
<div data-bbox="[181, 407, 816,... | 
# Hybrid Reasoning Agents
4 February 2024
## In-depth Analyses
RoadMap. We further analyze from 3 complementary perspectives to assess the method design and better understand its hybrid reasoning behavior. (i) T... | synthetic_layout_real_text | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18861 | ["table"] | {"font_size": 12, "font_format": "sans_serif", "table_style": "borderless", "figure_style": "bordered", "text_color": "dark_gray", "column_layout": "one_column"} | {} | [] |
synthetic_00000036 | 36 | full240_2026-10-14/sample_00042 | 2 | page-images/synthetic_00000036.png | <div data-bbox="[191, 160, 807, 402]" data-label="Table"><table><thead><tr><th>Method</th><th>AIME25</th><th>AIME26</th><th>MATH500</th><th>GPQA-D</th><th>MedQA</th><th>MuSiQue</th></tr></thead><tbody><tr><td>Coconut</td><td>13.1</td><td>9.6</td><td>51.4</td><td>26.8</td><td>61.7</td><td>14.0</td></tr><tr><td>SoftCoT</... | 
<table><tr><th>Method</th><th>AIME25</th><th>AIME26</th><th>MATH500</th><th>GPQA-D</th><th>MedQA</th><th>MuSiQue</th></tr><tr><td>Coconut</td><td>13.1</td><td>9.6</td><td>51.4</td><td>26.8</td><td>61.7</td><td>14.0</td></tr><tr><td>SoftCoT</td><td>16.5</td><td>17.9</td><td>63.6</td><td>28.9</t... | synthetic_layout_real_text | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18861 | ["table"] | {"font_size": 12, "font_format": "sans_serif", "table_style": "borderless", "figure_style": "bordered", "text_color": "dark_gray", "column_layout": "one_column"} | {} | [] |
synthetic_00000037 | 37 | full240_2026-10-14/sample_00043 | 1 | page-images/synthetic_00000037.png | <div data-bbox="[225, 209, 773, 279]" data-label="Section-Header"><h1>Post‑2024 Vaccination Campaign Failure Analysis at the Kalgoorlie Lithium Mine</h1></div>
<div data-bbox="[425, 333, 574, 343]" data-label="Text"><p>7 December 2001</p></div>
<div data-bbox="[221, 385, 636, 401]" data-label="Section-Header"><h2>Backg... | 
# Post‑2024 Vaccination Campaign Failure Analysis at the Kalgoorlie Lithium Mine
7 December 2001
## Background and Failure Overview
The onsite influenza vaccination drive for shift workers was scheduled for June... | synthetic_both | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18861 | ["bibliography", "code_block", "figure", "formula", "list", "page_footer", "page_header", "table"] | {"font_size": 10, "font_format": "monospace", "table_style": "striped", "figure_style": "bordered", "text_color": "black", "column_layout": "one_column"} | {"genre": "course syllabus", "register": "practical and instructional, addressing the reader directly", "angle": "an analysis of a failure and its causes", "setting": "a mining operation in Australia", "title_style": "a long specific title of 9 to 14 words naming the setting or year", "section_count": 2, "paragraphs_pe... | [] |
synthetic_00000038 | 38 | full240_2026-10-14/sample_00043 | 2 | page-images/synthetic_00000038.png | <div data-bbox="[325, 158, 674, 361]" data-label="Figure"><img alt="Flow diagram of vaccine distribution from depot to field teams."/></div>
<div data-bbox="[219, 377, 733, 400]" data-label="Caption"><p>Figure 1: Flow diagram of vaccine distribution from depot to field teams.</p></div>
<div data-bbox="[259, 422, 516, 4... | 
Flow diagram of vaccine distribution from depot to field teams.
Figure 1: Flow diagram of vaccine distribution from depot to field teams.
- Review scheduling software
- Upgrade temperature monitoring
- Add buffer stock
![heading](220,511... | synthetic_both | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18861 | ["bibliography", "code_block", "figure", "formula", "list", "page_footer", "page_header", "table"] | {"font_size": 10, "font_format": "monospace", "table_style": "striped", "figure_style": "bordered", "text_color": "black", "column_layout": "one_column"} | {"genre": "course syllabus", "register": "practical and instructional, addressing the reader directly", "angle": "an analysis of a failure and its causes", "setting": "a mining operation in Australia", "title_style": "a long specific title of 9 to 14 words naming the setting or year", "section_count": 2, "paragraphs_pe... | [] |
synthetic_00000039 | 39 | full240_2026-10-14/sample_00043 | 3 | page-images/synthetic_00000039.png | <div data-bbox="[219, 162, 759, 233]" data-label="Text"><p>2:1 ratio of field kits to cold‑chain boxes, and conducting quarterly simulation exercises. Adoption of the formula V = D / (T * C) where V is vaccination effectiveness, D is doses delivered, T is time in hours, and C is compliance factor, will help quantify pe... | 
2:1 ratio of field kits to cold‑chain boxes, and conducting quarterly simulation exercises. Adoption of the formula V = D / (T * C) where V is vaccination effectiveness, D is doses delivered, T is time in hours, and C is compliance factor, will help quantify performance in real time.
![code](2... | synthetic_both | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18861 | ["bibliography", "code_block", "figure", "formula", "list", "page_footer", "page_header", "table"] | {"font_size": 10, "font_format": "monospace", "table_style": "striped", "figure_style": "bordered", "text_color": "black", "column_layout": "one_column"} | {"genre": "course syllabus", "register": "practical and instructional, addressing the reader directly", "angle": "an analysis of a failure and its causes", "setting": "a mining operation in Australia", "title_style": "a long specific title of 9 to 14 words naming the setting or year", "section_count": 2, "paragraphs_pe... | [] |
synthetic_00000040 | 40 | full240_2026-10-14/sample_00046 | 1 | page-images/synthetic_00000040.png | <div data-bbox="[209, 220, 786, 293]" data-label="Section-Header"><h1>Visual Tokens on Demand: Visual Budgeting for Scalable and Efficient Multi-Image Reasoning in Multimodal LLMs</h1></div>
<div data-bbox="[449, 348, 548, 358]" data-label="Text"><p>2013-09-05</p></div>
<div data-bbox="[208, 405, 486, 416]" data-label=... | 
# Visual Tokens on Demand: Visual Budgeting for Scalable and Efficient Multi-Image Reasoning in Multimodal LLMs
2013-09-05
## Architecture Details
Each image is encoded by a frozen ImageNet-pretrained EfficientN... | synthetic_layout_real_text | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18858 | [] | {"font_size": 11, "font_format": "monospace", "table_style": "ruled", "figure_style": "borderless", "text_color": "dark_gray", "column_layout": "one_column"} | {} | [] |
synthetic_00000041 | 41 | full240_2026-10-14/sample_00047 | 1 | page-images/synthetic_00000041.png | <div data-bbox="[435, 197, 564, 216]" data-label="Section-Header"><h1>VoxPolicy</h1></div>
<div data-bbox="[440, 265, 559, 275]" data-label="Text"><p>27 June 2010</p></div>
<div data-bbox="[120, 307, 277, 323]" data-label="Section-Header"><h2>Background</h2></div>
<div data-bbox="[118, 339, 459, 393]" data-label="Text"... | 
# VoxPolicy
27 June 2010
## Background
The municipal health office launched a pilot to collect voice data from elderly speakers of the Xingu language in rural districts.
The aim was to ... | synthetic_both | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18858 | ["bibliography", "footnote", "formula", "page_header"] | {"font_size": 10, "font_format": "monospace", "table_style": "borderless", "figure_style": "bordered", "text_color": "dark_blue", "column_layout": "two_column"} | {"genre": "patient information leaflet", "register": "narrative, recounting events in order", "angle": "a review of a policy and its effects", "setting": "a municipal government office in Brazil", "title_style": "a short label of 1 to 3 words", "section_count": 8, "paragraphs_per_section": 3, "sentences_per_paragraph":... | [] |
synthetic_00000042 | 42 | full240_2026-10-14/sample_00047 | 2 | page-images/synthetic_00000042.png | <div data-bbox="[119, 159, 250, 170]" data-label="Section-Header"><h2>Outcomes</h2></div>
<div data-bbox="[118, 190, 475, 231]" data-label="Text"><p>By September 2024 the system achieved a recognition accuracy of 78 percent on test sentences from the Xingu speakers.</p></div>
<div data-bbox="[118, 235, 467, 292]" data-... | 
## Outcomes
By September 2024 the system achieved a recognition accuracy of 78 percent on test sentences from the Xingu speakers.
Healthcare staff reported faster entry of patient notes, improving appointment throughput by an estimated 10 t... | synthetic_both | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18858 | ["bibliography", "footnote", "formula", "page_header"] | {"font_size": 10, "font_format": "monospace", "table_style": "borderless", "figure_style": "bordered", "text_color": "dark_blue", "column_layout": "two_column"} | {"genre": "patient information leaflet", "register": "narrative, recounting events in order", "angle": "a review of a policy and its effects", "setting": "a municipal government office in Brazil", "title_style": "a short label of 1 to 3 words", "section_count": 8, "paragraphs_per_section": 3, "sentences_per_paragraph":... | [] |
synthetic_00000043 | 43 | full240_2026-10-14/sample_00051 | 1 | page-images/synthetic_00000043.png | <div data-bbox="[255, 217, 742, 237]" data-label="Section-Header"><h1>Checklist for Remote Learning Deployment</h1></div>
<div data-bbox="[438, 291, 560, 302]" data-label="Text"><p>22 March 2010</p></div>
<div data-bbox="[208, 348, 538, 364]" data-label="Section-Header"><h2>Implementation Checklist</h2></div>
<div data... | 
# Checklist for Remote Learning Deployment
22 March 2010
## Implementation Checklist
We prepared a concise checklist to guide the rapid rollout of the eLearning platform EduConnect v2.3 in Da Nang Regional Hospi... | synthetic_both | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18857 | ["bibliography", "footnote", "formula", "page_header", "table"] | {"font_size": 11, "font_format": "serif", "table_style": "striped", "figure_style": "bordered", "text_color": "dark_blue", "column_layout": "one_column"} | {"genre": "clinical case summary", "register": "conversational technical note, first person plural", "angle": "a checklist of requirements with brief justification", "setting": "a regional hospital in Vietnam", "title_style": "a plain descriptive title of 4 to 8 words", "section_count": 1, "paragraphs_per_section": 2, ... | [] |
synthetic_00000044 | 44 | full240_2026-10-14/sample_00051 | 2 | page-images/synthetic_00000044.png | <div data-bbox="[215, 166, 782, 270]" data-label="Table"><table><thead><tr><th>Metric</th><th>Target</th><th>Actual</th><th>Units</th><th>Notes</th></tr></thead><tbody><tr><td>Average bandwidth</td><td>3</td><td>2.8</td><td>Mbps</td><td>Slightly below target</td></tr><tr><td>Login success rate</td><td>95</td><td>97</td... | 
<table><tr><th>Metric</th><th>Target</th><th>Actual</th><th>Units</th><th>Notes</th></tr><tr><td>Average bandwidth</td><td>3</td><td>2.8</td><td>Mbps</td><td>Slightly below target</td></tr><tr><td>Login success rate</td><td>95</td><td>97</td><td>Percent</td><td>Meets requirement</td></tr><tr><... | synthetic_both | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18857 | ["bibliography", "footnote", "formula", "page_header", "table"] | {"font_size": 11, "font_format": "serif", "table_style": "striped", "figure_style": "bordered", "text_color": "dark_blue", "column_layout": "one_column"} | {"genre": "clinical case summary", "register": "conversational technical note, first person plural", "angle": "a checklist of requirements with brief justification", "setting": "a regional hospital in Vietnam", "title_style": "a plain descriptive title of 4 to 8 words", "section_count": 1, "paragraphs_per_section": 2, ... | [] |
synthetic_00000045 | 45 | full240_2026-10-14/sample_00055 | 1 | page-images/synthetic_00000045.png | <div data-bbox="[131, 205, 867, 275]" data-label="Section-Header"><h1>Senior NLP Engineer – Natural Language Inference Benchmark Development for Ho Chi Minh City Hospital 2024</h1></div>
<div data-bbox="[450, 334, 549, 344]" data-label="Text"><p>2022-11-26</p></div>
<div data-bbox="[120, 382, 363, 394]" data-label="Sec... | 
# Senior NLP Engineer – Natural Language Inference Benchmark Development for Ho Chi Minh City Hospital 2024
2022-11-26
## Position Overview
The Department of Clinical Informatics seeks a specialist to design, im... | synthetic_both | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18856 | ["bibliography", "figure", "formula", "page_header", "table"] | {"font_size": 11, "font_format": "monospace", "table_style": "striped", "figure_style": "borderless", "text_color": "dark_blue", "column_layout": "two_column"} | {"genre": "job description", "register": "formal, for specialists, dense with terminology", "angle": "a checklist of requirements with brief justification", "setting": "a regional hospital in Vietnam", "title_style": "a long specific title of 9 to 14 words naming the setting or year", "section_count": 4, "paragraphs_pe... | [] |
synthetic_00000046 | 46 | full240_2026-10-14/sample_00055 | 2 | page-images/synthetic_00000046.png | <div data-bbox="[118, 169, 473, 215]" data-label="Text"><p>validation, demonstrated by prior work that reports confidence intervals at the 95 percent level.</p></div>
<div data-bbox="[118, 244, 447, 258]" data-label="Section-Header"><h2>Preferred Qualifications</h2></div>
<div data-bbox="[118, 277, 482, 341]" data-labe... | 
validation, demonstrated by prior work that reports confidence intervals at the 95 percent level.
## Preferred Qualifications
Publication record in top‑tier NLP venues on domain‑specific benchmark creation, ideally with a focus on Asian lan... | synthetic_both | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18856 | ["bibliography", "figure", "formula", "page_header", "table"] | {"font_size": 11, "font_format": "monospace", "table_style": "striped", "figure_style": "borderless", "text_color": "dark_blue", "column_layout": "two_column"} | {"genre": "job description", "register": "formal, for specialists, dense with terminology", "angle": "a checklist of requirements with brief justification", "setting": "a regional hospital in Vietnam", "title_style": "a long specific title of 9 to 14 words naming the setting or year", "section_count": 4, "paragraphs_pe... | [] |
synthetic_00000047 | 47 | full240_2026-10-14/sample_00058 | 1 | page-images/synthetic_00000047.png | <div data-bbox="[205, 219, 794, 330]" data-label="Section-Header"><h1>Two Clocks: Why Larger Rollout Batches Learn Slower but Fail on Schedule in Reinforcement Learning with Verifiable Rewards</h1></div>
<div data-bbox="[417, 397, 582, 413]" data-label="Text"><p>18 August 2012</p></div>
<div data-bbox="[184, 455, 414, ... | 
# Two Clocks: Why Larger Rollout Batches Learn Slower but Fail on Schedule in Reinforcement Learning with Verifiable Rewards
18 August 2012
## Preliminaries
For each prompt, the policy samples G responses and re... | synthetic_layout_real_text | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18853 | [] | {"font_size": 12, "font_format": "monospace", "table_style": "ruled", "figure_style": "bordered", "text_color": "dark_gray", "column_layout": "one_column"} | {} | [] |
synthetic_00000048 | 48 | full240_2026-10-14/sample_00058 | 2 | page-images/synthetic_00000048.png | <div data-bbox="[182, 164, 763, 196]" data-label="Text"><p>accuracy after N responses and N() the first N at which it reaches a target .</p></div>
<div data-bbox="[182, 229, 731, 275]" data-label="Section-Header"><h2>Method: A Two-Clock Model of Batch Scaling</h2></div>
<div data-bbox="[181, 294, 808, 435]" data-label=... | 
accuracy after N responses and N() the first N at which it reaches a target .
## Method: A Two-Clock Model of Batch Scaling
We now formalize the two-clock picture. We model one optimizer step on a locally quadratic loss and ask which per-sa... | synthetic_layout_real_text | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18853 | [] | {"font_size": 12, "font_format": "monospace", "table_style": "ruled", "figure_style": "bordered", "text_color": "dark_gray", "column_layout": "one_column"} | {} | [] |
synthetic_00000049 | 49 | full240_2026-10-14/sample_00058 | 3 | page-images/synthetic_00000049.png | <div data-bbox="[181, 164, 815, 414]" data-label="Text"><p>We regress Nfail on b, with b the responses per optimizer step, and on , in two ways, and report both in Table 1. Least squares uses the eighteen runs that fail. A right-censored normal likelihood (Tobit) adds the three runs that never fail, whose budget is a l... | 
We regress Nfail on b, with b the responses per optimizer step, and on , in two ways, and report both in Table 1. Least squares uses the eighteen runs that fail. A right-censored normal likelihood (Tobit) adds the three runs that never fail, whose budget is a lower bound on their failure point.... | synthetic_layout_real_text | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18853 | [] | {"font_size": 12, "font_format": "monospace", "table_style": "ruled", "figure_style": "bordered", "text_color": "dark_gray", "column_layout": "one_column"} | {} | [] |
synthetic_00000050 | 50 | full240_2026-10-14/sample_00058 | 4 | page-images/synthetic_00000050.png | <div data-bbox="[181, 164, 813, 378]" data-label="Text"><p>expensive because this data must be regenerated as the policy changes: every policy update requires fresh rollouts, and generation dominates the cost. Practitioners therefore enlarge the rollout batch, the number of responses generated by one policy before it i... | 
expensive because this data must be regenerated as the policy changes: every policy update requires fresh rollouts, and generation dominates the cost. Practitioners therefore enlarge the rollout batch, the number of responses generated by one policy before it is updated, and rely on the square-... | synthetic_layout_real_text | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18853 | [] | {"font_size": 12, "font_format": "monospace", "table_style": "ruled", "figure_style": "bordered", "text_color": "dark_gray", "column_layout": "one_column"} | {} | [] |
synthetic_00000051 | 51 | full240_2026-10-14/sample_00059 | 1 | page-images/synthetic_00000051.png | <div data-bbox="[432, 204, 567, 223]" data-label="Section-Header"><h1>Refund Gap</h1></div>
<div data-bbox="[445, 277, 553, 291]" data-label="Text"><p>April 2, 2008</p></div>
<div data-bbox="[120, 328, 254, 341]" data-label="Section-Header"><h2>Abstract</h2></div>
<div data-bbox="[118, 362, 492, 700]" data-label="Text"... | 
# Refund Gap
April 2, 2008
## Abstract
This abstract examines the systematic failure of refund processing for defective canned pineapple sold by a Thai food processing plant on the e-commerce platform MarketPlac... | synthetic_both | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18853 | ["code_block", "list", "page_footer", "table"] | {"font_size": 11, "font_format": "serif", "table_style": "striped", "figure_style": "borderless", "text_color": "black", "column_layout": "two_column"} | {"genre": "conference abstract", "register": "persuasive, arguing for a recommendation", "angle": "an analysis of a failure and its causes", "setting": "a food processing plant in Thailand", "title_style": "a short label of 1 to 3 words", "section_count": 1, "paragraphs_per_section": 2, "sentences_per_paragraph": 5, "t... | [] |
synthetic_00000052 | 52 | full240_2026-10-14/sample_00062 | 1 | page-images/synthetic_00000052.png | <div data-bbox="[203, 205, 795, 252]" data-label="Section-Header"><h1>MFG: Multi-turn Feedback Guided Molecular Optimization</h1></div>
<div data-bbox="[425, 307, 574, 319]" data-label="Text"><p>January 6, 2005</p></div>
<div data-bbox="[120, 358, 251, 373]" data-label="Section-Header"><h2>Appendix</h2></div>
<div data... | 
# MFG: Multi-turn Feedback Guided Molecular Optimization
January 6, 2005
## Appendix
Computational methods for molecular optimization has been explored through a variety of paradigms, including molecule string t... | synthetic_layout_real_text | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18851 | [] | {"font_size": 11, "font_format": "monospace", "table_style": "ruled", "figure_style": "bordered", "text_color": "black", "column_layout": "two_column"} | {} | [] |
synthetic_00000053 | 53 | full240_2026-10-14/sample_00063 | 1 | page-images/synthetic_00000053.png | <div data-bbox="[228, 217, 770, 293]" data-label="Section-Header"><h1>Should the Municipal Government Approve the Expansion of the Brazilian Coffee Franchise 'Café Verde' into the City of Porto Alegre?</h1></div>
<div data-bbox="[455, 347, 542, 358]" data-label="Text"><p>2014-07-12</p></div>
<div data-bbox="[207, 401, ... | 
# Should the Municipal Government Approve the Expansion of the Brazilian Coffee Franchise 'Café Verde' into the City of Porto Alegre?
2014-07-12
## Course Description
This module provides an in-depth cost-benefi... | synthetic_both | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18851 | ["bibliography", "figure", "footnote"] | {"font_size": 11, "font_format": "sans_serif", "table_style": "borderless", "figure_style": "borderless", "text_color": "black", "column_layout": "one_column"} | {"genre": "course syllabus", "register": "formal, for specialists, dense with terminology", "angle": "a cost and benefit discussion", "setting": "a municipal government office in Brazil", "title_style": "a title phrased as a question", "section_count": 3, "paragraphs_per_section": 1, "sentences_per_paragraph": 5, "text... | [] |
synthetic_00000054 | 54 | full240_2026-10-14/sample_00063 | 2 | page-images/synthetic_00000054.png | <div data-bbox="[206, 168, 791, 266]" data-label="Text"><p>disruptions that could elevate coffee bean procurement costs by up to 7 percent. Students will practice drafting policy briefs that integrate stakeholder analysis, highlighting the positions of local business associations, consumer groups, and environmental NGO... | 
disruptions that could elevate coffee bean procurement costs by up to 7 percent. Students will practice drafting policy briefs that integrate stakeholder analysis, highlighting the positions of local business associations, consumer groups, and environmental NGOs. Finally, participants will deve... | synthetic_both | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18851 | ["bibliography", "figure", "footnote"] | {"font_size": 11, "font_format": "sans_serif", "table_style": "borderless", "figure_style": "borderless", "text_color": "black", "column_layout": "one_column"} | {"genre": "course syllabus", "register": "formal, for specialists, dense with terminology", "angle": "a cost and benefit discussion", "setting": "a municipal government office in Brazil", "title_style": "a title phrased as a question", "section_count": 3, "paragraphs_per_section": 1, "sentences_per_paragraph": 5, "text... | [] |
synthetic_00000055 | 55 | full240_2026-10-14/sample_00063 | 3 | page-images/synthetic_00000055.png | <div data-bbox="[323, 164, 674, 794]" data-label="Figure"><img alt="Projected cash flow diagram for Café Verde's flagship outlet over a five‑year horizon."/></div>
<div data-bbox="[206, 814, 790, 843]" data-label="Caption"><p>Figure 1: Projected cash flow diagram for Café Verde's flagship outlet over a five‑year horizo... | 
Projected cash flow diagram for Café Verde's flagship outlet over a five‑year horizon.
Figure 1: Projected cash flow diagram for Café Verde's flagship outlet over a five‑year horizon. | synthetic_both | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18851 | ["bibliography", "figure", "footnote"] | {"font_size": 11, "font_format": "sans_serif", "table_style": "borderless", "figure_style": "borderless", "text_color": "black", "column_layout": "one_column"} | {"genre": "course syllabus", "register": "formal, for specialists, dense with terminology", "angle": "a cost and benefit discussion", "setting": "a municipal government office in Brazil", "title_style": "a title phrased as a question", "section_count": 3, "paragraphs_per_section": 1, "sentences_per_paragraph": 5, "text... | [] |
synthetic_00000056 | 56 | full240_2026-10-14/sample_00063 | 4 | page-images/synthetic_00000056.png | <div data-bbox="[206, 165, 468, 178]" data-label="Section-Header"><h2>Assessment Methods</h2></div>
<div data-bbox="[206, 821, 791, 848]" data-label="Footnote"><small>All monetary values are expressed in Brazilian Reais and reflect estimates as of the 2026 fiscal year.</small></div>
<div data-bbox="[205, 199, 792, 366]... | 
## Assessment Methods
All monetary values are expressed in Brazilian Reais and reflect estimates as of the 2026 fiscal year.
Evaluation will consist of a mid‑term case study report (40 percent) requiring a detailed cash‑flow projection ... | synthetic_both | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18851 | ["bibliography", "figure", "footnote"] | {"font_size": 11, "font_format": "sans_serif", "table_style": "borderless", "figure_style": "borderless", "text_color": "black", "column_layout": "one_column"} | {"genre": "course syllabus", "register": "formal, for specialists, dense with terminology", "angle": "a cost and benefit discussion", "setting": "a municipal government office in Brazil", "title_style": "a title phrased as a question", "section_count": 3, "paragraphs_per_section": 1, "sentences_per_paragraph": 5, "text... | [] |
synthetic_00000057 | 57 | full240_2026-10-14/sample_00066 | 1 | page-images/synthetic_00000057.png | <div data-bbox="[161, 195, 837, 242]" data-label="Section-Header"><h1>WorldRecon: Generative Multi-View Mesh Reconstruction for Large-Scale 3D World</h1></div>
<div data-bbox="[423, 294, 576, 306]" data-label="Text"><p>November 27, 2011</p></div>
<div data-bbox="[119, 337, 377, 354]" data-label="Section-Header"><h2>Exp... | 
# WorldRecon: Generative Multi-View Mesh Reconstruction for Large-Scale 3D World
November 27, 2011
## Experimental Details
We use VGGT-() [3] prediction as the common geometric prior for all methods. It provides... | synthetic_layout_real_text | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18846 | [] | {"font_size": 10, "font_format": "sans_serif", "table_style": "ruled", "figure_style": "bordered", "text_color": "black", "column_layout": "two_column"} | {} | [] |
synthetic_00000058 | 58 | full240_2026-10-14/sample_00067 | 1 | page-images/synthetic_00000058.png | <div data-bbox="[278, 195, 723, 215]" data-label="Section-Header"><h1>Memory Safety in Central Bank Systems</h1></div>
<div data-bbox="[455, 266, 543, 277]" data-label="Text"><p>2022-06-08</p></div>
<div data-bbox="[119, 308, 285, 321]" data-label="Section-Header"><h2>Introduction</h2></div>
<div data-bbox="[118, 344, ... | 
# Memory Safety in Central Bank Systems
2022-06-08
## Introduction
The European Central Bank recently discovered a latent buffer overflow in its real‑time transaction monitoring module, prompting a comprehensive... | synthetic_both | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18846 | ["footnote", "formula", "page_footer", "page_header", "table"] | {"font_size": 10, "font_format": "sans_serif", "table_style": "striped", "figure_style": "borderless", "text_color": "dark_gray", "column_layout": "two_column"} | {"genre": "cooking and nutrition guide", "register": "narrative, recounting events in order", "angle": "a safety and risk discussion", "setting": "a central bank in the Eurozone", "title_style": "a plain descriptive title of 4 to 8 words", "section_count": 7, "paragraphs_per_section": 3, "sentences_per_paragraph": 1, "... | [] |
synthetic_00000059 | 59 | full240_2026-10-14/sample_00067 | 2 | page-images/synthetic_00000059.png | <div data-bbox="[118, 161, 491, 204]" data-label="Text"><p>performance overhead, and the compliance with regulatory standards such as GDPR (General Data Protection Regulation).</p></div>
<div data-bbox="[127, 221, 482, 297]" data-label="Table"><table><thead><tr><th>Issue Type</th><th>Count</th><th>Patch Applied</th><th... | 
performance overhead, and the compliance with regulatory standards such as GDPR (General Data Protection Regulation).
<table><tr><th>Issue Type</th><th>Count</th><th>Patch Applied</th><th>Residual Risk</th></tr><tr><td>Buffer Overflow</td><td>12</td><td>Yes</td><td>Lo... | synthetic_both | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18846 | ["footnote", "formula", "page_footer", "page_header", "table"] | {"font_size": 10, "font_format": "sans_serif", "table_style": "striped", "figure_style": "borderless", "text_color": "dark_gray", "column_layout": "two_column"} | {"genre": "cooking and nutrition guide", "register": "narrative, recounting events in order", "angle": "a safety and risk discussion", "setting": "a central bank in the Eurozone", "title_style": "a plain descriptive title of 4 to 8 words", "section_count": 7, "paragraphs_per_section": 3, "sentences_per_paragraph": 1, "... | [] |
synthetic_00000060 | 60 | full240_2026-10-14/sample_00070 | 1 | page-images/synthetic_00000060.png | <div data-bbox="[215, 217, 783, 304]" data-label="Section-Header"><h1>The Last Remote Sensing Model Built by Humans? Casting Any Earth Observation Task as Recursive Self-Improvement</h1></div>
<div data-bbox="[446, 366, 551, 379]" data-label="Text"><p>2015-11-28</p></div>
<div data-bbox="[183, 431, 399, 446]" data-labe... | 
# The Last Remote Sensing Model Built by Humans? Casting Any Earth Observation Task as Recursive Self-Improvement
2015-11-28
## Introduction
Earth observation (EO) models are still built by hand. Each new sensor... | synthetic_layout_real_text | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18845 | [] | {"font_size": 12, "font_format": "serif", "table_style": "striped", "figure_style": "borderless", "text_color": "dark_blue", "column_layout": "one_column"} | {} | [] |
synthetic_00000061 | 61 | full240_2026-10-14/sample_00070 | 2 | page-images/synthetic_00000061.png | <div data-bbox="[181, 163, 816, 414]" data-label="Text"><p>Our premise is that any verifiable task can be cast as a harness and a verifier. The harness defines what the agent may use, such as data, pretrained checkpoints, libraries and compute, and the verifier defines what counts as success by scoring the agent's mode... | 
Our premise is that any verifiable task can be cast as a harness and a verifier. The harness defines what the agent may use, such as data, pretrained checkpoints, libraries and compute, and the verifier defines what counts as success by scoring the agent's model ( 1). Every EO task can be made ... | synthetic_layout_real_text | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18845 | [] | {"font_size": 12, "font_format": "serif", "table_style": "striped", "figure_style": "borderless", "text_color": "dark_blue", "column_layout": "one_column"} | {} | [] |
synthetic_00000062 | 62 | full240_2026-10-14/sample_00070 | 3 | page-images/synthetic_00000062.png | <div data-bbox="[182, 163, 816, 214]" data-label="Text"><p>methods. The third is how the amount of labeled data in the harness and the choice of agent change the outcome. 1 then examines the runs themselves to explain where the accuracy comes from.</p></div> | 
methods. The third is how the amount of labeled data in the harness and the choice of agent change the outcome. 1 then examines the runs themselves to explain where the accuracy comes from. | synthetic_layout_real_text | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18845 | [] | {"font_size": 12, "font_format": "serif", "table_style": "striped", "figure_style": "borderless", "text_color": "dark_blue", "column_layout": "one_column"} | {} | [] |
synthetic_00000063 | 63 | full240_2026-10-14/sample_00071 | 1 | page-images/synthetic_00000063.png | <div data-bbox="[240, 209, 760, 279]" data-label="Section-Header"><h1>2025 Quarterly Forecast for Handloom Textile Production at MetroTransit Mexico</h1></div>
<div data-bbox="[430, 333, 569, 346]" data-label="Text"><p>3 January 2021</p></div>
<div data-bbox="[221, 385, 463, 401]" data-label="Section-Header"><h2>Execut... | 
# 2025 Quarterly Forecast for Handloom Textile Production at MetroTransit Mexico
3 January 2021
## Executive Summary
We anticipate a 4.5 percent increase in output of the traditional Talavera handloom line for t... | synthetic_both | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18845 | ["bibliography", "figure", "footnote", "formula"] | {"font_size": 10, "font_format": "monospace", "table_style": "striped", "figure_style": "borderless", "text_color": "dark_blue", "column_layout": "one_column"} | {"genre": "financial quarterly report", "register": "conversational technical note, first person plural", "angle": "a forecast with stated assumptions", "setting": "a public transit authority in Mexico", "title_style": "a title starting with a number or a year", "section_count": 4, "paragraphs_per_section": 1, "sentenc... | [] |
synthetic_00000064 | 64 | full240_2026-10-14/sample_00071 | 2 | page-images/synthetic_00000064.png | <div data-bbox="[331, 162, 668, 376]" data-label="Figure"><img alt="Projected production volume versus actual output for Q1 2025."/></div>
<div data-bbox="[219, 396, 776, 419]" data-label="Caption"><p>Figure 1: Projected production volume versus actual output for Q1 2025.</p></div>
<div data-bbox="[219, 449, 474, 465]"... | 
Projected production volume versus actual output for Q1 2025.
Figure 1: Projected production volume versus actual output for Q1 2025.
## Operational Impact
The upgraded shuttle‑loom equipment, delivering 1.8 m... | synthetic_both | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18845 | ["bibliography", "figure", "footnote", "formula"] | {"font_size": 10, "font_format": "monospace", "table_style": "striped", "figure_style": "borderless", "text_color": "dark_blue", "column_layout": "one_column"} | {"genre": "financial quarterly report", "register": "conversational technical note, first person plural", "angle": "a forecast with stated assumptions", "setting": "a public transit authority in Mexico", "title_style": "a title starting with a number or a year", "section_count": 4, "paragraphs_per_section": 1, "sentenc... | [] |
synthetic_00000065 | 65 | full240_2026-10-14/sample_00071 | 3 | page-images/synthetic_00000065.png | <div data-bbox="[222, 161, 728, 186]" data-label="Text"><p>[4] Diaz. Forecasting Methods for Small Scale Manufacturing. 2020.</p></div> | 
[4] Diaz. Forecasting Methods for Small Scale Manufacturing. 2020. | synthetic_both | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18845 | ["bibliography", "figure", "footnote", "formula"] | {"font_size": 10, "font_format": "monospace", "table_style": "striped", "figure_style": "borderless", "text_color": "dark_blue", "column_layout": "one_column"} | {"genre": "financial quarterly report", "register": "conversational technical note, first person plural", "angle": "a forecast with stated assumptions", "setting": "a public transit authority in Mexico", "title_style": "a title starting with a number or a year", "section_count": 4, "paragraphs_per_section": 1, "sentenc... | [] |
synthetic_00000066 | 66 | full240_2026-10-14/sample_00075 | 1 | page-images/synthetic_00000066.png | <div data-bbox="[137, 195, 862, 242]" data-label="Section-Header"><h1>Pedestrian Flow Assessment: Impact of Seasonal Market Stalls on the Historic Centre of Cape Town</h1></div>
<div data-bbox="[455, 294, 543, 304]" data-label="Text"><p>2004-11-23</p></div>
<div data-bbox="[119, 336, 285, 349]" data-label="Section-Head... | 
# Pedestrian Flow Assessment: Impact of Seasonal Market Stalls on the Historic Centre of Cape Town
2004-11-23
## Introduction
The City of Cape Town Municipal Planning Department commissioned a short-term measure... | synthetic_both | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18844 | ["code_block", "page_footer", "page_header", "table"] | {"font_size": 10, "font_format": "sans_serif", "table_style": "striped", "figure_style": "borderless", "text_color": "dark_gray", "column_layout": "two_column"} | {"genre": "agricultural extension bulletin", "register": "formal, for specialists, dense with terminology", "angle": "a measurement study with concrete results", "setting": "a legal practice in South Africa", "title_style": "a title with a colon and a subtitle", "section_count": 4, "paragraphs_per_section": 2, "sentenc... | [] |
synthetic_00000067 | 67 | full240_2026-10-14/sample_00078 | 1 | page-images/synthetic_00000067.png | <div data-bbox="[247, 209, 751, 256]" data-label="Section-Header"><h1>Who Improves the Robot? Toward Physical Recursive Self-Improvement</h1></div>
<div data-bbox="[450, 305, 549, 315]" data-label="Text"><p>2019-07-30</p></div>
<div data-bbox="[221, 358, 671, 373]" data-label="Section-Header"><h2>Harness Adaptation of ... | 
# Who Improves the Robot? Toward Physical Recursive Self-Improvement
2019-07-30
## Harness Adaptation of Physical RSI
The task execution loop connects planning, execution, observation, and correction (Figure. 1)... | synthetic_layout_real_text | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18842 | [] | {"font_size": 10, "font_format": "monospace", "table_style": "borderless", "figure_style": "borderless", "text_color": "dark_gray", "column_layout": "one_column"} | {} | [] |
synthetic_00000068 | 68 | full240_2026-10-14/sample_00078 | 2 | page-images/synthetic_00000068.png | <div data-bbox="[220, 159, 671, 174]" data-label="Section-Header"><h2>Safety and Control of Physical RSI</h2></div>
<div data-bbox="[219, 190, 776, 306]" data-label="Text"><p>Physical RSI must preserve safety as both the task-execution system M=(,H) and the improvement procedure evolve. System updates can change which ... | 
## Safety and Control of Physical RSI
Physical RSI must preserve safety as both the task-execution system M=(,H) and the improvement procedure evolve. System updates can change which states the system visits and alter its hazard estimates and action interfaces. Envi... | synthetic_layout_real_text | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18842 | [] | {"font_size": 10, "font_format": "monospace", "table_style": "borderless", "figure_style": "borderless", "text_color": "dark_gray", "column_layout": "one_column"} | {} | [] |
synthetic_00000069 | 69 | full240_2026-10-14/sample_00079 | 1 | page-images/synthetic_00000069.png | <div data-bbox="[260, 209, 738, 256]" data-label="Section-Header"><h1>Will Our Real-Time Leak Detection Module Meet the SLA Requirements?</h1></div>
<div data-bbox="[420, 305, 578, 318]" data-label="Text"><p>8 September 2021</p></div>
<div data-bbox="[221, 352, 511, 367]" data-label="Section-Header"><h2>Assumptions and... | 
# Will Our Real-Time Leak Detection Module Meet the SLA Requirements?
8 September 2021
## Assumptions and Scope
We assume that the municipal water network operates under a constant pressure regime and that senso... | synthetic_both | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18842 | ["bibliography", "code_block", "formula", "list", "page_footer"] | {"font_size": 10, "font_format": "monospace", "table_style": "ruled", "figure_style": "borderless", "text_color": "dark_blue", "column_layout": "one_column"} | {"genre": "legal contract clause commentary", "register": "conversational technical note, first person plural", "angle": "a forecast with stated assumptions", "setting": "a software startup in Canada", "title_style": "a title phrased as a question", "section_count": 4, "paragraphs_per_section": 2, "sentences_per_paragr... | [] |
synthetic_00000070 | 70 | full240_2026-10-14/sample_00079 | 2 | page-images/synthetic_00000070.png | <div data-bbox="[219, 159, 550, 174]" data-label="Section-Header"><h2>Compliance and Reporting</h2></div>
<div data-bbox="[219, 190, 765, 231]" data-label="Text"><p>All data handling will comply with the Personal Information Protection and Electronic Documents Act (PIPEDA) and the Ontario Water Act, with encrypted stor... | 
## Compliance and Reporting
All data handling will comply with the Personal Information Protection and Electronic Documents Act (PIPEDA) and the Ontario Water Act, with encrypted storage at rest.
Monthly compliance reports will be generated... | synthetic_both | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18842 | ["bibliography", "code_block", "formula", "list", "page_footer"] | {"font_size": 10, "font_format": "monospace", "table_style": "ruled", "figure_style": "borderless", "text_color": "dark_blue", "column_layout": "one_column"} | {"genre": "legal contract clause commentary", "register": "conversational technical note, first person plural", "angle": "a forecast with stated assumptions", "setting": "a software startup in Canada", "title_style": "a title phrased as a question", "section_count": 4, "paragraphs_per_section": 2, "sentences_per_paragr... | [] |
synthetic_00000071 | 71 | full240_2026-10-14/sample_00083 | 1 | page-images/synthetic_00000071.png | <div data-bbox="[158, 202, 842, 225]" data-label="Section-Header"><h1>GPU Kernel Optimizer: Release 3.2.1 Patch Notes</h1></div>
<div data-bbox="[410, 289, 589, 305]" data-label="Text"><p>February 15, 2018</p></div>
<div data-bbox="[120, 344, 489, 364]" data-label="Section-Header"><h2>Requirements Checklist</h2></div>
... | 
# GPU Kernel Optimizer: Release 3.2.1 Patch Notes
February 15, 2018
## Requirements Checklist
All checks were performed on the Nvidia Ampere architecture using driver version 525.89.12.
![text](118,382,491,... | synthetic_both | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18839 | ["footnote", "list", "page_header"] | {"font_size": 12, "font_format": "serif", "table_style": "ruled", "figure_style": "bordered", "text_color": "dark_gray", "column_layout": "two_column"} | {"genre": "software release notes", "register": "terse bulletin style, fragments allowed where natural", "angle": "a checklist of requirements with brief justification", "setting": "no particular place, kept generic", "title_style": "a title with a colon and a subtitle", "section_count": 1, "paragraphs_per_section": 1,... | [] |
synthetic_00000072 | 72 | full240_2026-10-14/sample_00086 | 1 | page-images/synthetic_00000072.png | <div data-bbox="[138, 204, 862, 223]" data-label="Section-Header"><h1>Scaling Online Optimal Transport Minibatches for Flow Matching</h1></div>
<div data-bbox="[455, 278, 543, 289]" data-label="Text"><p>2001-08-28</p></div>
<div data-bbox="[119, 325, 293, 338]" data-label="Section-Header"><h2>Main Results</h2></div>
<d... | 
# Scaling Online Optimal Transport Minibatches for Flow Matching
2001-08-28
## Main Results
We first study the benefit of scaling OT minibatch size through its effect on stochastic-gradient variance. The closest... | synthetic_layout_real_text | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18838 | [] | {"font_size": 11, "font_format": "sans_serif", "table_style": "borderless", "figure_style": "bordered", "text_color": "dark_gray", "column_layout": "two_column"} | {} | [] |
synthetic_00000073 | 73 | full240_2026-10-14/sample_00087 | 1 | page-images/synthetic_00000073.png | <div data-bbox="[437, 207, 562, 223]" data-label="Section-Header"><h1>Lane Audit</h1></div>
<div data-bbox="[420, 278, 578, 292]" data-label="Text"><p>November 10, 2001</p></div>
<div data-bbox="[221, 330, 392, 347]" data-label="Section-Header"><h2>Background</h2></div>
<div data-bbox="[219, 361, 780, 388]" data-label=... | 
# Lane Audit
November 10, 2001
## Background
The Greenfield Secondary School in Lagos has recently installed a 500 metre bicycle lane alongside its main driveway.
Local students use the... | synthetic_both | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18838 | ["code_block", "footnote", "formula", "page_header", "table"] | {"font_size": 10, "font_format": "serif", "table_style": "borderless", "figure_style": "bordered", "text_color": "black", "column_layout": "one_column"} | {"genre": "agricultural extension bulletin", "register": "journalistic, with concrete examples and quotes", "angle": "a cost and benefit discussion", "setting": "a secondary school in Nigeria", "title_style": "a short label of 1 to 3 words", "section_count": 6, "paragraphs_per_section": 3, "sentences_per_paragraph": 1,... | [] |
synthetic_00000074 | 74 | full240_2026-10-14/sample_00087 | 2 | page-images/synthetic_00000074.png | <div data-bbox="[272, 159, 725, 247]" data-label="Table"><table><thead><tr><th>Item</th><th>Cost (NGN)</th><th>Savings (NGN)</th><th>Net (NGN)</th></tr></thead><tbody><tr><td>Reflective posts</td><td>100000</td><td>0</td><td>100000</td></tr><tr><td>Signage</td><td>20000</td><td>0</td><td>20000</td></tr><tr><td>Safety o... | 
<table><tr><th>Item</th><th>Cost (NGN)</th><th>Savings (NGN)</th><th>Net (NGN)</th></tr><tr><td>Reflective posts</td><td>100000</td><td>0</td><td>100000</td></tr><tr><td>Signage</td><td>20000</td><td>0</td><td>20000</td></tr><tr><td>Safety officer</td><td>36000</td><td>0</td><td>36000</td></tr... | synthetic_both | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18838 | ["code_block", "footnote", "formula", "page_header", "table"] | {"font_size": 10, "font_format": "serif", "table_style": "borderless", "figure_style": "bordered", "text_color": "black", "column_layout": "one_column"} | {"genre": "agricultural extension bulletin", "register": "journalistic, with concrete examples and quotes", "angle": "a cost and benefit discussion", "setting": "a secondary school in Nigeria", "title_style": "a short label of 1 to 3 words", "section_count": 6, "paragraphs_per_section": 3, "sentences_per_paragraph": 1,... | [] |
synthetic_00000075 | 75 | full240_2026-10-14/sample_00090 | 1 | page-images/synthetic_00000075.png | <div data-bbox="[186, 219, 811, 299]" data-label="Section-Header"><h1>Plan-Bench: Can Existing VLMs Go Well Beyond Perception to Reason and Plan in Embodied Worlds?</h1></div>
<div data-bbox="[440, 366, 558, 378]" data-label="Text"><p>2016-01-13</p></div>
<div data-bbox="[184, 430, 488, 444]" data-label="Section-Header... | 
# Plan-Bench: Can Existing VLMs Go Well Beyond Perception to Reason and Plan in Embodied Worlds?
2016-01-13
## Additional Results
Tables 1-- 1 report model accuracy on all 16 subtasks of Plan-Bench, revealing di... | synthetic_layout_real_text | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18836 | [] | {"font_size": 12, "font_format": "monospace", "table_style": "ruled", "figure_style": "borderless", "text_color": "dark_gray", "column_layout": "one_column"} | {} | [] |
synthetic_00000076 | 76 | full240_2026-10-14/sample_00090 | 2 | page-images/synthetic_00000076.png | <div data-bbox="[182, 160, 608, 174]" data-label="Section-Header"><h2>Limitations and Discussion</h2></div>
<div data-bbox="[181, 198, 805, 280]" data-label="Text"><p>We construct Plan-Bench to evaluate the reasoning and planning capabilities of VLMs in embodied tasks across Spatial Cognition, Temporal Reasoning, and M... | 
## Limitations and Discussion
We construct Plan-Bench to evaluate the reasoning and planning capabilities of VLMs in embodied tasks across Spatial Cognition, Temporal Reasoning, and Multi-step Planning, while several limitations remain to be further explored and add... | synthetic_layout_real_text | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18836 | [] | {"font_size": 12, "font_format": "monospace", "table_style": "ruled", "figure_style": "borderless", "text_color": "dark_gray", "column_layout": "one_column"} | {} | [] |
synthetic_00000077 | 77 | full240_2026-10-14/sample_00091 | 1 | page-images/synthetic_00000077.png | <div data-bbox="[232, 217, 765, 261]" data-label="Section-Header"><h1>1978 Implementation of the SEIR Mining Health Model</h1></div>
<div data-bbox="[455, 319, 542, 330]" data-label="Text"><p>2002-02-22</p></div>
<div data-bbox="[207, 372, 368, 388]" data-label="Section-Header"><h2>Background</h2></div>
<div data-bbox=... | 
# 1978 Implementation of the SEIR Mining Health Model
2002-02-22
## Background
In 1975 the Westfield coal mine introduced the SEIR-Mine v2.1 framework to quantify respiratory disease spread among underground wor... | synthetic_both | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18836 | ["code_block", "figure", "list", "page_footer", "table"] | {"font_size": 11, "font_format": "sans_serif", "table_style": "striped", "figure_style": "bordered", "text_color": "black", "column_layout": "one_column"} | {"genre": "standard operating procedure", "register": "formal, for specialists, dense with terminology", "angle": "a historical account of how the practice developed", "setting": "a mining operation in Australia", "title_style": "a title starting with a number or a year", "section_count": 3, "paragraphs_per_section": 2... | [] |
synthetic_00000078 | 78 | full240_2026-10-14/sample_00091 | 2 | page-images/synthetic_00000078.png | <div data-bbox="[228, 168, 553, 294]" data-label="List-Group"><ol><li>Define shift boundaries</li><li>Collect aerosol samples</li><li>Run SEIR-Mine v2.1</li><li>Log compartment counts</li><li>Activate mitigation if threshold exceeded</li></ol></div>
<div data-bbox="[206, 323, 349, 336]" data-label="Section-Header"><h2>... | 
1. Define shift boundaries
2. Collect aerosol samples
3. Run SEIR-Mine v2.1
4. Log compartment counts
5. Activate mitigation if threshold exceeded
## Outcomes
The 1978 rollout reduced projected outbreak peaks by 37 percent, confirming the u... | synthetic_both | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18836 | ["code_block", "figure", "list", "page_footer", "table"] | {"font_size": 11, "font_format": "sans_serif", "table_style": "striped", "figure_style": "bordered", "text_color": "black", "column_layout": "one_column"} | {"genre": "standard operating procedure", "register": "formal, for specialists, dense with terminology", "angle": "a historical account of how the practice developed", "setting": "a mining operation in Australia", "title_style": "a title starting with a number or a year", "section_count": 3, "paragraphs_per_section": 2... | [] |
synthetic_00000079 | 79 | full240_2026-10-14/sample_00094 | 1 | page-images/synthetic_00000079.png | <div data-bbox="[230, 217, 768, 265]" data-label="Section-Header"><h1>RoboSimStudio: An Interactive Toolkit for Rapid Real2Sim of Robot Manipulation Tasks</h1></div>
<div data-bbox="[419, 319, 578, 333]" data-label="Text"><p>September 15, 2014</p></div>
<div data-bbox="[207, 367, 379, 381]" data-label="Section-Header">... | 
# RoboSimStudio: An Interactive Toolkit for Rapid Real2Sim of Robot Manipulation Tasks
September 15, 2014
## Case Studies
We first study human-guided Real2Sim using visual feedback. Starting from observations of... | synthetic_layout_real_text | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18835 | [] | {"font_size": 11, "font_format": "sans_serif", "table_style": "ruled", "figure_style": "borderless", "text_color": "dark_blue", "column_layout": "one_column"} | {} | [] |
synthetic_00000080 | 80 | full240_2026-10-14/sample_00094 | 2 | page-images/synthetic_00000080.png | <div data-bbox="[206, 168, 791, 249]" data-label="Text"><p>contact and forces. composes a scene with robots, objects, and physics into one simulation while keeping modeling choices, including the solver, contact model, and physical parameters, explicit and editable. The composed scene can be accessed through code, an i... | 
contact and forces. composes a scene with robots, objects, and physics into one simulation while keeping modeling choices, including the solver, contact model, and physical parameters, explicit and editable. The composed scene can be accessed through code, an interactive studio for a human, or ... | synthetic_layout_real_text | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18835 | [] | {"font_size": 11, "font_format": "sans_serif", "table_style": "ruled", "figure_style": "borderless", "text_color": "dark_blue", "column_layout": "one_column"} | {} | [] |
synthetic_00000081 | 81 | full240_2026-10-14/sample_00094 | 3 | page-images/synthetic_00000081.png | <div data-bbox="[206, 168, 791, 318]" data-label="Text"><p>We recognize that both humans and LLM coding agents share a fundamental interaction pattern for Real2Sim: they modify a scene, execute physical interactions, observe discrepancies, and refine the model based on feedback. Specifically, humans perform this loop t... | 
We recognize that both humans and LLM coding agents share a fundamental interaction pattern for Real2Sim: they modify a scene, execute physical interactions, observe discrepancies, and refine the model based on feedback. Specifically, humans perform this loop through direct manipulation and imm... | synthetic_layout_real_text | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18835 | [] | {"font_size": 11, "font_format": "sans_serif", "table_style": "ruled", "figure_style": "borderless", "text_color": "dark_blue", "column_layout": "one_column"} | {} | [] |
synthetic_00000082 | 82 | full240_2026-10-14/sample_00098 | 1 | page-images/synthetic_00000082.png | <div data-bbox="[183, 217, 815, 272]" data-label="Section-Header"><h1>Game2Real-Bench: Do Toy Games Reveal and Transfer to Real-World Visual Reasoning?</h1></div>
<div data-bbox="[446, 335, 552, 348]" data-label="Text"><p>2021-11-24</p></div>
<div data-bbox="[183, 399, 693, 419]" data-label="Section-Header"><h2>Detaile... | 
# Game2Real-Bench: Do Toy Games Reveal and Transfer to Real-World Visual Reasoning?
2021-11-24
## Detailed Benchmark Specification
We describe the game tasks in Game2Real-Bench and the accompanying Game2Real-Dat... | synthetic_layout_real_text | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18834 | [] | {"font_size": 12, "font_format": "serif", "table_style": "striped", "figure_style": "borderless", "text_color": "black", "column_layout": "one_column"} | {} | [] |
synthetic_00000083 | 83 | full240_2026-10-14/sample_00098 | 2 | page-images/synthetic_00000083.png | <div data-bbox="[181, 163, 816, 323]" data-label="Text"><p>recognizing what is present in an image, visual reasoning requires models to understand how objects are related, how constraints interact, and how actions change a scene. These demands arise in everyday tasks, such as interpreting an assembly step, as well as i... | 
recognizing what is present in an image, visual reasoning requires models to understand how objects are related, how constraints interact, and how actions change a scene. These demands arise in everyday tasks, such as interpreting an assembly step, as well as in toy games with explicit rules an... | synthetic_layout_real_text | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18834 | [] | {"font_size": 12, "font_format": "serif", "table_style": "striped", "figure_style": "borderless", "text_color": "black", "column_layout": "one_column"} | {} | [] |
synthetic_00000084 | 84 | full240_2026-10-14/sample_00102 | 1 | page-images/synthetic_00000084.png | <div data-bbox="[225, 217, 773, 265]" data-label="Section-Header"><h1>OmniCompass: A Holistic Benchmark for Omni-Modal Video Grounding and Understanding</h1></div>
<div data-bbox="[434, 319, 564, 333]" data-label="Text"><p>15 January 2009</p></div>
<div data-bbox="[207, 376, 506, 389]" data-label="Section-Header"><h2>B... | 
# OmniCompass: A Holistic Benchmark for Omni-Modal Video Grounding and Understanding
15 January 2009
## Benchmark Construction
OmniCompass-VL contains 580 selected public-source videos and 240 self-collected cli... | synthetic_layout_real_text | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18829 | [] | {"font_size": 11, "font_format": "sans_serif", "table_style": "ruled", "figure_style": "borderless", "text_color": "dark_gray", "column_layout": "one_column"} | {} | [] |
synthetic_00000085 | 85 | full240_2026-10-14/sample_00102 | 2 | page-images/synthetic_00000085.png | <div data-bbox="[206, 165, 342, 181]" data-label="Section-Header"><h2>Appendix</h2></div>
<div data-bbox="[206, 199, 791, 229]" data-label="Text"><p>Appendixomnicompass-appendix [2] #1 apx#2 #1#2 tocdepth2 linkcolor=black black @starttocapx</p></div> | 
## Appendix
Appendixomnicompass-appendix [2] #1 apx#2 #1#2 tocdepth2 linkcolor=black black @starttocapx | synthetic_layout_real_text | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18829 | [] | {"font_size": 11, "font_format": "sans_serif", "table_style": "ruled", "figure_style": "borderless", "text_color": "dark_gray", "column_layout": "one_column"} | {} | [] |
synthetic_00000086 | 86 | full240_2026-10-14/sample_00106 | 1 | page-images/synthetic_00000086.png | <div data-bbox="[276, 209, 722, 283]" data-label="Section-Header"><h1>WorldOracle: Benchmarking Action-Conditioned World Models against Real-World Executions</h1></div>
<div data-bbox="[430, 333, 569, 346]" data-label="Text"><p>23 August 2023</p></div>
<div data-bbox="[221, 385, 353, 397]" data-label="Section-Header"><... | 
# WorldOracle: Benchmarking Action-Conditioned World Models against Real-World Executions
23 August 2023
## Overview
WorldOracle: Benchmarking Action-Conditioned World Models against Real-World Executions | synthetic_layout_real_text | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18826 | [] | {"font_size": 10, "font_format": "monospace", "table_style": "striped", "figure_style": "borderless", "text_color": "dark_gray", "column_layout": "one_column"} | {} | [] |
synthetic_00000087 | 87 | full240_2026-10-14/sample_00118 | 1 | page-images/synthetic_00000087.png | <div data-bbox="[268, 209, 730, 283]" data-label="Section-Header"><h1>DuplexBridge: Coordinating Understanding and Generation for Unified Image Editing</h1></div>
<div data-bbox="[440, 333, 558, 343]" data-label="Text"><p>6 March 2011</p></div>
<div data-bbox="[221, 385, 573, 397]" data-label="Section-Header"><h2>Discu... | 
# DuplexBridge: Coordinating Understanding and Generation for Unified Image Editing
6 March 2011
## Discussion and Limitations
DuplexBridge improves joint task fulfillment by separating local supervision referen... | synthetic_layout_real_text | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18819 | [] | {"font_size": 10, "font_format": "monospace", "table_style": "borderless", "figure_style": "bordered", "text_color": "dark_gray", "column_layout": "one_column"} | {} | [] |
synthetic_00000088 | 88 | full240_2026-10-14/sample_00118 | 2 | page-images/synthetic_00000088.png | <div data-bbox="[220, 159, 697, 193]" data-label="Section-Header"><h2>Coordinating Learning with Local and Global Feedback</h2></div>
<div data-bbox="[219, 212, 777, 283]" data-label="Text"><p>DuplexBridge separates three evaluations by the information they receive: prompt fidelity uses (X,u,), execution uses (X,, Y), ... | 
## Coordinating Learning with Local and Global Feedback
DuplexBridge separates three evaluations by the information they receive: prompt fidelity uses (X,u,), execution uses (X,, Y), and the global outcome uses (X,u, Y) ( 1). Local rewards train each role's responsi... | synthetic_layout_real_text | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18819 | [] | {"font_size": 10, "font_format": "monospace", "table_style": "borderless", "figure_style": "bordered", "text_color": "dark_gray", "column_layout": "one_column"} | {} | [] |
synthetic_00000089 | 89 | full240_2026-10-14/sample_00118 | 3 | page-images/synthetic_00000089.png | <div data-bbox="[219, 159, 536, 174]" data-label="Section-Header"><h2>Qualitative Comparisons</h2></div>
<div data-bbox="[219, 190, 388, 200]" data-label="Text"><p>type=figure,skip=5pt</p></div>
<div data-bbox="[219, 205, 777, 276]" data-label="Text"><p>Qualitative comparison on RISEBench. Rows cover untying a knot, a ... | 
## Qualitative Comparisons
type=figure,skip=5pt
Qualitative comparison on RISEBench. Rows cover untying a knot, a visual equivalence relation, maze navigation, a bitten apple, and cooking progression. The maze requires a connected path thro... | synthetic_layout_real_text | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18819 | [] | {"font_size": 10, "font_format": "monospace", "table_style": "borderless", "figure_style": "bordered", "text_color": "dark_gray", "column_layout": "one_column"} | {} | [] |
synthetic_00000090 | 90 | full240_2026-10-14/sample_00122 | 1 | page-images/synthetic_00000090.png | <div data-bbox="[194, 217, 805, 299]" data-label="Section-Header"><h1>Back-Translation Still Matters: Leveraging Monolingual Data for Low-Resource Machine Translation with LLMs</h1></div>
<div data-bbox="[428, 367, 569, 383]" data-label="Text"><p>April 16, 2013</p></div>
<div data-bbox="[183, 425, 815, 468]" data-label... | 
# Back-Translation Still Matters: Leveraging Monolingual Data for Low-Resource Machine Translation with LLMs
April 16, 2013
## Languages and Backbones: Full Flores-200 Results
This section gives the Flores-200 [... | synthetic_layout_real_text | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18818 | [] | {"font_size": 12, "font_format": "serif", "table_style": "ruled", "figure_style": "borderless", "text_color": "dark_gray", "column_layout": "one_column"} | {} | [] |
synthetic_00000091 | 91 | full240_2026-10-14/sample_00122 | 2 | page-images/synthetic_00000091.png | <div data-bbox="[181, 163, 816, 287]" data-label="Text"><p>Our question is whether an LLM with a weak prior can leverage monolingual text for better translation, and how that text should enter training. The recipes in 1 differ in what the model learns to predict. CPT predicts original text but learns no mapping between... | 
Our question is whether an LLM with a weak prior can leverage monolingual text for better translation, and how that text should enter training. The recipes in 1 differ in what the model learns to predict. CPT predicts original text but learns no mapping between languages. Self-distillation lear... | synthetic_layout_real_text | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18818 | [] | {"font_size": 12, "font_format": "serif", "table_style": "ruled", "figure_style": "borderless", "text_color": "dark_gray", "column_layout": "one_column"} | {} | [] |
synthetic_00000092 | 92 | full240_2026-10-14/sample_00126 | 1 | page-images/synthetic_00000092.png | <div data-bbox="[345, 204, 654, 223]" data-label="Section-Header"><h1>Self-Evolving Agent Harness</h1></div>
<div data-bbox="[438, 277, 561, 290]" data-label="Text"><p>March 23, 2025</p></div>
<div data-bbox="[119, 327, 268, 340]" data-label="Section-Header"><h2>Conclusion</h2></div>
<div data-bbox="[118, 361, 490, 476... | 
# Self-Evolving Agent Harness
March 23, 2025
## Conclusion
Our framework defines harness self-evolution through feedback-driven revision, continuing update capability, and persistence around fixed models. It sep... | synthetic_layout_real_text | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18813 | [] | {"font_size": 11, "font_format": "sans_serif", "table_style": "ruled", "figure_style": "borderless", "text_color": "dark_blue", "column_layout": "two_column"} | {} | [] |
synthetic_00000093 | 93 | full240_2026-10-14/sample_00130 | 1 | page-images/synthetic_00000093.png | <div data-bbox="[221, 195, 777, 215]" data-label="Section-Header"><h1>Statistical Vignettes on the Grokking Phenomenon</h1></div>
<div data-bbox="[455, 266, 543, 277]" data-label="Text"><p>2008-05-15</p></div>
<div data-bbox="[119, 308, 444, 321]" data-label="Section-Header"><h2>Proofs of the Main Results</h2></div>
<d... | 
# Statistical Vignettes on the Grokking Phenomenon
2008-05-15
## Proofs of the Main Results
Throughout, convergence and stochastic order estimates are with respect to the training sample and initialization. We u... | synthetic_layout_real_text | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18811 | [] | {"font_size": 10, "font_format": "sans_serif", "table_style": "ruled", "figure_style": "bordered", "text_color": "dark_gray", "column_layout": "two_column"} | {} | [] |
synthetic_00000094 | 94 | full240_2026-10-14/sample_00142 | 1 | page-images/synthetic_00000094.png | <div data-bbox="[251, 220, 747, 293]" data-label="Section-Header"><h1>Registration is a Representation Problem: Solving Whole-Body PET/CT Registration</h1></div>
<div data-bbox="[449, 348, 548, 358]" data-label="Text"><p>2001-10-10</p></div>
<div data-bbox="[208, 405, 326, 416]" data-label="Section-Header"><h2>Methods<... | 
# Registration is a Representation Problem: Solving Whole-Body PET/CT Registration
2001-10-10
## Methods
Our two-stage pipeline is visualized in Figure 1. Stage 1 uses the images and their nonlinear representati... | synthetic_layout_real_text | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18804 | [] | {"font_size": 11, "font_format": "monospace", "table_style": "borderless", "figure_style": "borderless", "text_color": "dark_gray", "column_layout": "one_column"} | {} | [] |
synthetic_00000095 | 95 | full240_2026-10-14/sample_00142 | 2 | page-images/synthetic_00000095.png | <div data-bbox="[205, 169, 788, 523]" data-label="Text"><p>Multi-stage iterative optimization. Given the representations defined above for both moving and fixed volumes, we take a hybrid approach and use an off-the-shelf GPU-accelerated registration solver (FireANTs [5, 6]) to optimize a deformation field. We take a tw... | 
Multi-stage iterative optimization. Given the representations defined above for both moving and fixed volumes, we take a hybrid approach and use an off-the-shelf GPU-accelerated registration solver (FireANTs [5, 6]) to optimize a deformation field. We take a two-stage registration approach. Sta... | synthetic_layout_real_text | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18804 | [] | {"font_size": 11, "font_format": "monospace", "table_style": "borderless", "figure_style": "borderless", "text_color": "dark_gray", "column_layout": "one_column"} | {} | [] |
synthetic_00000096 | 96 | full240_2026-10-14/sample_00142 | 3 | page-images/synthetic_00000096.png | <div data-bbox="[206, 166, 375, 182]" data-label="Section-Header"><h2>Experiments</h2></div>
<div data-bbox="[205, 200, 778, 348]" data-label="Text"><p>Dataset. We participate in the 2026 Learn2Reg PSMAReg Challenge. The pairs are longitudinal whole-body PSMA PET/CT scans that are downsampled by the challenge organizer... | 
## Experiments
Dataset. We participate in the 2026 Learn2Reg PSMAReg Challenge. The pairs are longitudinal whole-body PSMA PET/CT scans that are downsampled by the challenge organizers to 192192288 at 2.73442.73443.27 mm resolution. The training set contains 202 ima... | synthetic_layout_real_text | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18804 | [] | {"font_size": 11, "font_format": "monospace", "table_style": "borderless", "figure_style": "borderless", "text_color": "dark_gray", "column_layout": "one_column"} | {} | [] |
synthetic_00000097 | 97 | full240_2026-10-14/sample_00146 | 1 | page-images/synthetic_00000097.png | <div data-bbox="[239, 219, 757, 238]" data-label="Section-Header"><h1>SEAR: Self-Evolving Agentic Robotics</h1></div>
<div data-bbox="[449, 293, 548, 302]" data-label="Text"><p>2022-08-15</p></div>
<div data-bbox="[208, 349, 701, 384]" data-label="Section-Header"><h2>Auto-Research: Supply, Time, and Task Structure</h2>... | 
# SEAR: Self-Evolving Agentic Robotics
2022-08-15
## Auto-Research: Supply, Time, and Task Structure
These tables support 1. The groups are observational, not randomized interventions. A RoboLab session records ... | synthetic_layout_real_text | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18792 | [] | {"font_size": 11, "font_format": "monospace", "table_style": "ruled", "figure_style": "bordered", "text_color": "dark_blue", "column_layout": "one_column"} | {} | [] |
synthetic_00000098 | 98 | full240_2026-10-14/sample_00146 | 2 | page-images/synthetic_00000098.png | <div data-bbox="[207, 166, 474, 182]" data-label="Section-Header"><h2>Interaction Budgets</h2></div>
<div data-bbox="[205, 199, 789, 346]" data-label="Text"><p>Per-task budgets. Robosuite, LIBERO, LIBERO-PRO, and LIBERO-RIG tasks receive a one-hour development budget, while RoboLab tasks receive two hours to accommodat... | 
## Interaction Budgets
Per-task budgets. Robosuite, LIBERO, LIBERO-PRO, and LIBERO-RIG tasks receive a one-hour development budget, while RoboLab tasks receive two hours to accommodate longer simulation times. Each run permits up to eight agent-initiated submissions... | synthetic_layout_real_text | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18792 | [] | {"font_size": 11, "font_format": "monospace", "table_style": "ruled", "figure_style": "bordered", "text_color": "dark_blue", "column_layout": "one_column"} | {} | [] |
synthetic_00000099 | 99 | full240_2026-10-14/sample_00146 | 3 | page-images/synthetic_00000099.png | <div data-bbox="[205, 169, 784, 386]" data-label="Text"><p>next round. Seen this way, the persistent state of the whole system plays the role that weights play elsewhere, except that much of it is readable, editable knowledge. Our results point at one architecture for that loop: the LM as a deliberative layer that plan... | 
next round. Seen this way, the persistent state of the whole system plays the role that weights play elsewhere, except that much of it is readable, editable knowledge. Our results point at one architecture for that loop: the LM as a deliberative layer that plans, diagnoses, and writes code, and... | synthetic_layout_real_text | synctex | ocr-layout-html-v2.1 | ocr-layout-md-v1 | true | synthetic-latex-pipeline | 2610.18792 | [] | {"font_size": 11, "font_format": "monospace", "table_style": "ruled", "figure_style": "bordered", "text_color": "dark_blue", "column_layout": "one_column"} | {} | [] |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.