Datasets:
image imagewidth (px) 1.02k 6.8k | image_name stringlengths 20 119 | prompt stringclasses 1
value | text stringlengths 40 32.2k | data_source stringclasses 7
values | language stringclasses 1
value | layout stringclasses 5
values | has_watermark bool 1
class | fuzzy_scan bool 1
class | width int32 1.02k 6.8k | height int32 1.24k 9.07k |
|---|---|---|---|---|---|---|---|---|---|---|
page-f9583127-1277-4424-be9d-35430e0aedd6.png | Extract the text from the above document as if you were reading it naturally. Return the tables in html format. Return the equations in LaTeX representation. If there is an image in the document wrap it as follows <img>image here</img>. Watermarks should be wrapped in brackets. Ex: <watermark>OFFICIAL COPY</watermark>.... | RWKV-7 uses four low-rank MLPs for decay $ w $ , value residual $ v $, in-context learning rate $ a $ and gate $ g $ respectively. The intermediate dimensions are listed in Table 16. These values are based on our mere speculation of how much information can be passed through.
<table><tr><td>Dimension $(D)$</td><td>$d_... | academic_literature | english | single_column | false | false | 1,700 | 2,200 | |
page-bb98165a-8033-4ea4-a527-58ac974f2c32.png | Extract the text from the above document as if you were reading it naturally. Return the tables in html format. Return the equations in LaTeX representation. If there is an image in the document wrap it as follows <img>image here</img>. Watermarks should be wrapped in brackets. Ex: <watermark>OFFICIAL COPY</watermark>.... | SEC. 5.3
INTERPOLATIONBY SPLINE FUNCTIONS
<page_number>291</page_number>
Figure 5.15 The parabolically terminated cubic spline.
Figure 5.16 The curvature adjusted cubic spline with $S ^ { \prime \prime } ( 0 ) = 0.3$and $ { S } ^ { \prime \prime } ( 3 ) = 3 . 3$.
and the solution is $ {m}_{1} = {2.7} $ and $ {m}_{... | book | english | single_column | false | false | 1,653 | 2,339 | |
page-9477c155-19b3-4f2d-ac18-ee3f8997d2f8.png | Extract the text from the above document as if you were reading it naturally. Return the tables in html format. Return the equations in LaTeX representation. If there is an image in the document wrap it as follows <img>image here</img>. Watermarks should be wrapped in brackets. Ex: <watermark>OFFICIAL COPY</watermark>.... | CAI MITRA WANG WIBISONO
#### G.1.3. BOUNDS ON THE SECOND MOMENT ALONG THE FINITE-PARTICLE SYSTEMS
In this section, we show that along the finite-particle dynamics (16) and algorithm (20), the distributions remain in $ \mathcal{P}\left( {\mathbb{R}}^{2dN}\right) $ .
Under Assumption 1, recall from Theorem 35 (in Sect... | academic_literature | english | single_column | false | false | 1,700 | 2,200 | |
page-1ce383cc-63a3-43c9-b6f2-7498a952b65c.png | Extract the text from the above document as if you were reading it naturally. Return the tables in html format. Return the equations in LaTeX representation. If there is an image in the document wrap it as follows <img>image here</img>. Watermarks should be wrapped in brackets. Ex: <watermark>OFFICIAL COPY</watermark>.... | <page_number>23</page_number>
$\bigcup_{{\left( {i,k}\right) \in {\mathcal{C}}_{n}^{1}}}{C}_{n,k}^{i} $ and $ {A}_{2} = \bigcup_{{\left( {i,k}\right) \in {\mathcal{C}}_{n}^{2}}}{C}_{n,k}^{i} $ with finite sets $ {\mathcal{C}}_{n}^{1} $ and $ {\mathcal{C}}_{n}^{2} $ of indices such that $ {A}_{1} \subseteq A \subseteq ... | academic_literature | english | single_column | false | false | 1,700 | 2,200 | |
page-f108020f-edc5-4a27-85d9-d9d9c5d67309.png | Extract the text from the above document as if you were reading it naturally. Return the tables in html format. Return the equations in LaTeX representation. If there is an image in the document wrap it as follows <img>image here</img>. Watermarks should be wrapped in brackets. Ex: <watermark>OFFICIAL COPY</watermark>.... | PROOF. We have $ M{M}^{\mathrm{H}} - M{N}^{\mathrm{H}} - N{M}^{\mathrm{H}} + N{N}^{\mathrm{H}} = \left( {M - N}\right) {\left( M - N\right) }^{\mathrm{H}} \succeq 0 $ and hence, (44) holds. $\square$
The following proposition provides a sufficient condition linear in $ {\bar{Z}}_{\mathrm{u}} $ which, ensures the exp... | academic_literature | english | double_column | false | false | 1,700 | 2,200 | |
page-dc728bd5-50a9-41e8-9e5a-d4b01e6664a2.png | Extract the text from the above document as if you were reading it naturally. Return the tables in html format. Return the equations in LaTeX representation. If there is an image in the document wrap it as follows <img>image here</img>. Watermarks should be wrapped in brackets. Ex: <watermark>OFFICIAL COPY</watermark>.... | <page_number>56</page_number>
X.CEN AND Z. SONG
We claim that
$$
{\int }_{Q}\left| {g\lambda }\right| \mathop{\prod }\limits_{{i \in \tau }}{\left( \mathop{\sum }\limits_{{P \in \tilde{\mathcal{S}},P \subseteq Q}}{\eta }_{i,P}{\chi }_{P}\left( x\right) \right) }^{{k}_{i} - {t}_{i}} \lesssim {\int }_{Q}{A}_{\tild... | academic_literature | english | single_column | false | false | 1,700 | 2,200 | |
page-d2816bbe-3c28-4905-ad6d-5f873b525de3.png | Extract the text from the above document as if you were reading it naturally. Return the tables in html format. Return the equations in LaTeX representation. If there is an image in the document wrap it as follows <img>image here</img>. Watermarks should be wrapped in brackets. Ex: <watermark>OFFICIAL COPY</watermark>.... | $$
\stackrel{\left( i\right) }{ \lesssim }{\|\mathbf{U}{\mathbf{U}}^{\top }{\mathbf{U}}^{ * }\|}_{2,\infty }\|{\mathbf{E}{\mathbf{V}}^{ * }}\|\|{\mathbf{\Lambda }}^{* - 1}\| + {\|\mathbf{R}\mathbf{V}{\mathbf{V}}^{\top }{\mathbf{V}}^{ * } - \mathbf{R}{\mathbf{V}}^{ * }\|}_{2,\infty }\|{\mathbf{\Lambda }}^{* - 1}\| + {\... | academic_literature | english | single_column | false | false | 1,700 | 2,200 | |
page-7a7566fd-43d8-46e4-8a4f-0330e1f27a88.png | Extract the text from the above document as if you were reading it naturally. Return the tables in html format. Return the equations in LaTeX representation. If there is an image in the document wrap it as follows <img>image here</img>. Watermarks should be wrapped in brackets. Ex: <watermark>OFFICIAL COPY</watermark>.... | Proof. For any $ {z}_{1},{z}_{2} \in \mathbb{C} $ , let $ {z}^{\theta } = \left( {1 - \theta }\right) {z}_{1} + \theta {z}_{2} $ and let $ \gamma \left( \theta \right) = f\left( {\left| {z}^{\theta }\right| }^{2}\right) {z}^{\theta } $ for $ 0 \leq \theta \leq 1 $ , we have
$$
f\left( {\left| {z}_{2}\right| }^{2}\righ... | academic_literature | english | single_column | false | false | 1,654 | 2,339 | |
page-d9a27df2-127e-4f76-bb61-e7e741078bef.png | Extract the text from the above document as if you were reading it naturally. Return the tables in html format. Return the equations in LaTeX representation. If there is an image in the document wrap it as follows <img>image here</img>. Watermarks should be wrapped in brackets. Ex: <watermark>OFFICIAL COPY</watermark>.... | K-MODULI OF PURE STATES OF FOUR QUBITS
<page_number>25</page_number>
where $ \varpi $ is the blow up of the lines $ \left\{ {{z}_{0} - {z}_{2} = {z}_{3} - {z}_{1} = 0}\right\} $ and $ \left\{ {{z}_{0} + {z}_{2} = {z}_{1} + {z}_{3} = 0}\right\} $ , and $ \varphi $ is a $ {\mathbb{P}}^{1} $ -bundle. Moreover,the map $ ... | academic_literature | english | single_column | false | false | 1,700 | 2,200 | |
page-7d2d833d-37b1-4c18-9ee6-4e09d352e265.png | Extract the text from the above document as if you were reading it naturally. Return the tables in html format. Return the equations in LaTeX representation. If there is an image in the document wrap it as follows <img>image here</img>. Watermarks should be wrapped in brackets. Ex: <watermark>OFFICIAL COPY</watermark>.... | <page_number>26</page_number>
S. GAUTAM AND M. T. MOHAN
Let us now consider
$$
\left|\int_0^T\langle\mathrm{B}_n(\boldsymbol{u}_n(t)),\mathrm{P}_n\boldsymbol{v}(t)\rangle\mathrm{d}t-\int_0^T\langle\mathrm{B}(\boldsymbol{u}(t)),\boldsymbol{v}(t)\rangle\mathrm{d}t\right|
$$
$$
\leq\left|\int_{0}^{T}\langle\mathrm{B}(... | academic_literature | english | single_column | false | false | 1,700 | 2,200 | |
page-83a675ab-0021-4979-8293-8af0d36146a0.png | Extract the text from the above document as if you were reading it naturally. Return the tables in html format. Return the equations in LaTeX representation. If there is an image in the document wrap it as follows <img>image here</img>. Watermarks should be wrapped in brackets. Ex: <watermark>OFFICIAL COPY</watermark>.... | DIFFERENTIABLE GROUPOID OBJECTS
<page_number>41</page_number>
Spelling out the actions $ {\beta }_{{T}^{n}E} $ and $ {\beta }_{{T}^{n}F} $ as given in (4.12), we get that the inner square is the outer rectangle of the following diagram
<img></img>
which commutes since $ \Phi $ is $ G $-equivariant and $ {T}^{n} $ i... | academic_literature | english | single_column | false | false | 1,700 | 2,200 | |
page-3ec8e45c-1c38-41d2-bb9d-c50745232f2a.png | Extract the text from the above document as if you were reading it naturally. Return the tables in html format. Return the equations in LaTeX representation. If there is an image in the document wrap it as follows <img>image here</img>. Watermarks should be wrapped in brackets. Ex: <watermark>OFFICIAL COPY</watermark>.... | which was required.
Completeness, Soundness and Entanglement lower bound: Here, we can assume that $\mathcal { V } $ is $\lambda$-bounded. Thus, by Claim 4.49, $\mathcal { V } _ { n } ^ { ( 1 ) } = \mathfrak { Padding } ( \mathcal { V } _ { n } , n ^ { \lambda } )$. Then, by Claim 4.50,$\mathcal { V } _ { n } ^ { ( 2 ... | book | english | single_column | false | false | 1,700 | 2,200 | |
page-b2cc5729-a6b2-49ba-af62-788bf069b617.png | Extract the text from the above document as if you were reading it naturally. Return the tables in html format. Return the equations in LaTeX representation. If there is an image in the document wrap it as follows <img>image here</img>. Watermarks should be wrapped in brackets. Ex: <watermark>OFFICIAL COPY</watermark>.... | <page_number>40</page_number>
$$
= O(s_n^2 \chi^m) + 4 \sum_{q_1, q_2 = 1}^{p} \Gamma_0^{q_1, q_2}(i/n, u)
$$
$$
\times \mathbb{E}(\epsilon_{i-k}^{r_1}(u) \sum_{l=(i-s_n) \vee 1}^{i-M-1} K_\tau(l/n - t)\check{\phi}_{l,k}^{r_1,q_1}(u) \epsilon_{i-k}^{r_2}(u) \sum_{l'=(i-s_n) \vee 1}^{i-M-1} K_\tau(l^{\prime}/n - t) \c... | book | english | single_column | false | false | 1,700 | 2,200 | |
page-9cca07f9-2e06-4da7-b4f4-496c6e305543.png | Extract the text from the above document as if you were reading it naturally. Return the tables in html format. Return the equations in LaTeX representation. If there is an image in the document wrap it as follows <img>image here</img>. Watermarks should be wrapped in brackets. Ex: <watermark>OFFICIAL COPY</watermark>.... | ON THE CONVERGENCE OF MIN-MAX LANGEVIN DYNAMICS AND ALGORITHM
Then we can compute:
$$
\frac{\partial \rho_t^Z}{\partial t} = \frac{\partial}{\partial t} \int_{\mathbb{R}^{2dN}} \rho_0^Z(z_0) \rho_{t|0}^Z(\cdot \mid z_0) \, dz_0
$$
$$
= \int_{\mathbb{R}^{2dN}} \rho_0^Z(z_0) \frac{\partial \rho_{t|0}^Z(\cdot \mid z_0)... | book | english | single_column | false | false | 1,700 | 2,200 | |
page-d209d5c0-fe1e-4cf0-90d8-6d610948b235.png | Extract the text from the above document as if you were reading it naturally. Return the tables in html format. Return the equations in LaTeX representation. If there is an image in the document wrap it as follows <img>image here</img>. Watermarks should be wrapped in brackets. Ex: <watermark>OFFICIAL COPY</watermark>.... | <page_number>50</page_number>
SOPHIE RAYNOR
Pairs $ \left( {{\mathcal{X}}^{1},{f}^{1}}\right) $ and $ \left( {{\mathcal{X}}^{1},{f}^{2}}\right) $ represent the same element $ {\lbrack \mathcal{X},f\rbrack }_{ * } \in {( \widetilde{L}{T}_{ * }{\Upsilon }_{ * }\mathcal{H}) }_{X} $ if and only if there is an object $ {\... | book | english | single_column | false | false | 1,700 | 2,200 | |
page-617783f8-bbd6-4faf-bca6-a05222c29314.png | Extract the text from the above document as if you were reading it naturally. Return the tables in html format. Return the equations in LaTeX representation. If there is an image in the document wrap it as follows <img>image here</img>. Watermarks should be wrapped in brackets. Ex: <watermark>OFFICIAL COPY</watermark>.... | A Nearly Optimal Single Loop Algorithm for Stochastic Bilevel Optimization under Unbounded Smoothness
where we define $ {v}_{t} $ as the following:
$$
{v}_{t} = \left\lbrack {{\nabla }_{yy}^{2}g\left( {{x}_{t},{y}_{t}}\right) - {\nabla }_{yy}^{2}G\left( {{x}_{t},{y}_{t};{\zeta }_{t}}\right) }\right\rbrack {z}_{t} ... | book | english | single_column | false | false | 1,700 | 2,200 | |
page-57d5842c-85c2-41c9-9d37-65aad59c1277.png | Extract the text from the above document as if you were reading it naturally. Return the tables in html format. Return the equations in LaTeX representation. If there is an image in the document wrap it as follows <img>image here</img>. Watermarks should be wrapped in brackets. Ex: <watermark>OFFICIAL COPY</watermark>.... | <page_number>8</page_number>
SANIA ASIF${}^{1}$, ZHIXIANG WU${}^{2}$
Definition 3.1. Let $ \left( {{\mathfrak{M}}_{\phi },\rho }\right) $ be a representation of an averaging Lie conformal algebra $ \left( {\mathfrak{L},{\left\lbrack \cdot {}_{\lambda} \cdot \right\rbrack }_{P}}\right) $ . Define a map $ {\xi }^{p} :... | book | english | single_column | false | false | 1,653 | 2,339 | |
page-404e0471-bd71-426a-8e5a-81b8327f88e5.png | Extract the text from the above document as if you were reading it naturally. Return the tables in html format. Return the equations in LaTeX representation. If there is an image in the document wrap it as follows <img>image here</img>. Watermarks should be wrapped in brackets. Ex: <watermark>OFFICIAL COPY</watermark>.... | $$
Z{Z}^{\top }\stackrel{\left( 1\right) }{ = }T\left\lbrack \begin{array}{cc} {D}_{\text{des }}^{\frac{1}{2}\top }\left( {\left( {\mathop{\sum }\limits_{{i = 1}}^{T}\bar{x}{\left( {\omega }_{i}\right) }^{\mathrm{H}}\bar{x}\left( {\omega }_{i}\right) }\right) \otimes {I}_{{n}_{\phi }}}\right) {D}_{\text{des }}^{\fra... | academic_literature | english | 1andmore_column | false | false | 1,700 | 2,200 | |
page-dad0f4e5-290f-496f-bbdd-099ad75c6ff0.png | Extract the text from the above document as if you were reading it naturally. Return the tables in html format. Return the equations in LaTeX representation. If there is an image in the document wrap it as follows <img>image here</img>. Watermarks should be wrapped in brackets. Ex: <watermark>OFFICIAL COPY</watermark>.... | HOMEOMORPHISMS OF THE PSEUDOARC
<page_number>23</page_number>
hence $ p{ \sqcap }_{{G}_{m}}q $ , as $ {\sqsupseteq}_{n}^{m} $ is $ \sqcap $ -preserving. Conversely, if $p{ \sqcap }_{{G}_{m}}q$ then we have $ r \in {G}_{m + 1} $ with $ p,q > r $ , as $\sqsupseteq_{{m + 1}}^{m} $ is edge-witnessing, showing that $ { \s... | academic_literature | english | single_column | false | false | 1,653 | 2,339 | |
page-a976a26e-4dbb-492a-a380-b923b87ed496.png | Extract the text from the above document as if you were reading it naturally. Return the tables in html format. Return the equations in LaTeX representation. If there is an image in the document wrap it as follows <img>image here</img>. Watermarks should be wrapped in brackets. Ex: <watermark>OFFICIAL COPY</watermark>.... | iff
$$
\left\lbrack\begin{array}{cccccc}{R}_{11} & \cdots & {R}_{1k} & \cdots & {R}_{1{t}_{j}}\\ \vdots & \ddots & \vdots & \vdots & \vdots\\ {R}_{k1} & \cdots & {R}_{kk} & \cdots & {R}_{kt_{j}}\\ \vdots & \vdots & \vdots & \ddots & \vdots\\ {R}_{{t}_{j}1} & \cdots & {R}_{{t}_{j}k} & \cdots & {R}_{{t}_{j}{t}_{j}}\end{... | book | english | single_column | false | false | 1,654 | 2,339 | |
page-39f38730-0ae6-4cfe-8b99-c65d06d09544.png | Extract the text from the above document as if you were reading it naturally. Return the tables in html format. Return the equations in LaTeX representation. If there is an image in the document wrap it as follows <img>image here</img>. Watermarks should be wrapped in brackets. Ex: <watermark>OFFICIAL COPY</watermark>.... | AN INFINITE HIERARCHY OF HÖLDER CUSPS FOR 1D EULER
<page_number>29</page_number>
for all $ t \in \left\lbrack {0,{T}_{ * }}\right) ,0 \leq m \leq {2n} + 1 $ .
Proposition 5.5. Suppose the hypotheses of Proposition 5.4 are satisfied,and additionally $ {\overline {C}}_{x} $ is a constant satisfying
$$
{\overline{C}}_... | academic_literature | english | single_column | false | false | 1,700 | 2,200 | |
page-4009495f-1a7c-46dc-9f52-7f80f5f0cc5c.png | Extract the text from the above document as if you were reading it naturally. Return the tables in html format. Return the equations in LaTeX representation. If there is an image in the document wrap it as follows <img>image here</img>. Watermarks should be wrapped in brackets. Ex: <watermark>OFFICIAL COPY</watermark>.... | <page_number>38</page_number>
ISAAC NEAL, STEVE SHKOLLER, AND VLAD VICOL
$$
+ \left( {m + 1}\right) q{\left\lbrack \frac{{\left( m + 1\right) }^{2}{\|{\Sigma }^{-{\delta m}}{\partial }_{t}^{m}{\mathring{K}}_{\lambda }\|}_{{L}_{x}^{q}}}{m!M{C}_{t}^{m}}\right\rbrack }^{q - 1}\lbrack {\frac{19}{9}{\overline{C}}_{t} +... | academic_literature | english | single_column | false | false | 1,700 | 2,200 | |
page-8524b54d-6372-4f9e-a662-5254349cfda3.png | Extract the text from the above document as if you were reading it naturally. Return the tables in html format. Return the equations in LaTeX representation. If there is an image in the document wrap it as follows <img>image here</img>. Watermarks should be wrapped in brackets. Ex: <watermark>OFFICIAL COPY</watermark>.... | $$
+\mathcal{L}[\mu](rf_n,r\underline{f}_n,r\overset{\circ}{{\lambda}}_n)+O(r^{-1}),
$$
$$
r^2\left(\operatorname{tr}\chi^\mathbf{S}-\frac{2}{r^\mathbf{S}}\right)=r^2\left(\operatorname{tr}\chi-\frac{2}{r}\right)+r\operatorname{div}^\mathbf{S}(rf_n)-2R_n^\#+rf_n\cdot r\eta+O(r^{-1}),
$$
$$
{r}^{2}\left( {\operatornam... | academic_literature | english | single_column | false | false | 1,654 | 2,339 | |
page-2861127b-41a4-44d5-b026-2de9cb408347.png | Extract the text from the above document as if you were reading it naturally. Return the tables in html format. Return the equations in LaTeX representation. If there is an image in the document wrap it as follows <img>image here</img>. Watermarks should be wrapped in brackets. Ex: <watermark>OFFICIAL COPY</watermark>.... | Virtual classes, real sets
2. David Makinson, Sets, Logic and Maths for Computing (Springer, 3rd edn 2020), Chapters 1 to 3.
Chapter 1 reviews basic facts about sets. Chapter 2 is on relations. Chapter 3 is on functions. This too can be warmly recommended - though you might want to supplement it by following up the r... | academic_literature | english | single_column | false | false | 1,338 | 1,922 | |
page-d127acc0-f8c9-4f01-8e87-b4c19be7cbb0.png | Extract the text from the above document as if you were reading it naturally. Return the tables in html format. Return the equations in LaTeX representation. If there is an image in the document wrap it as follows <img>image here</img>. Watermarks should be wrapped in brackets. Ex: <watermark>OFFICIAL COPY</watermark>.... | PROFILES
# THE MASTER BUILDER
Norman Foster's empire of image control.
BY IAN PARKER
<img></img>
good at designing. He's also very good at making his colleagues not stop designing.
Norman Foster, the British architect, resembles the titans he serves. His expansionist ambition and personal wealth set him alongside... | magazine | english | double_column | false | false | 1,575 | 2,150 | |
page-2dc37c8b-1547-43dd-8643-fdbecfe5eea8.png | Extract the text from the above document as if you were reading it naturally. Return the tables in html format. Return the equations in LaTeX representation. If there is an image in the document wrap it as follows <img>image here</img>. Watermarks should be wrapped in brackets. Ex: <watermark>OFFICIAL COPY</watermark>.... | ```
"#没有定义
[root@ubuntu ~]# ansible-playbook when-1.yaml
PLAY [localhost]
*
TASK [debug]
*
ok: [localhost] => {
"msg": "undefined"
}
PLAY RECAP
*
localhost : ok=1 changed=0 unreachable=0 failed=0 skipped=0 rescued=0 ignored=0
#定义
[root@ubuntu ~]# ansible-playbook -e tom=cat when-1.yaml
PLAY [localhost]
*
TASK [... | book | english | single_column | false | false | 1,653 | 2,339 | |
page-5ef91820-15a4-41af-b57f-36ec76879fee.png | Extract the text from the above document as if you were reading it naturally. Return the tables in html format. Return the equations in LaTeX representation. If there is an image in the document wrap it as follows <img>image here</img>. Watermarks should be wrapped in brackets. Ex: <watermark>OFFICIAL COPY</watermark>.... | <page_number>106</page_number>
PLATFORMS
combine the business dimension with the technical one at the large scale of permanent change and modularity. Whereas application domains like digital humanitarians are beginning to follow microservices, the Airbnb platform already deploys 3,500 microservices per week, with a t... | academic_literature | english | single_column | false | false | 1,229 | 1,843 | |
page-83587fc3-fe55-4e51-a3c5-f0664edba743.png | Extract the text from the above document as if you were reading it naturally. Return the tables in html format. Return the equations in LaTeX representation. If there is an image in the document wrap it as follows <img>image here</img>. Watermarks should be wrapped in brackets. Ex: <watermark>OFFICIAL COPY</watermark>.... | Proceedings of the Thirty-Second International Joint Conference on Artificial Intelligence (IJCAI-23) Sister Conferences Best Papers Track
<table><tr><td>Name</td><td>postE</td><td>Learned Invariant</td></tr><tr><td>Bin1</td><td>$n$</td><td>$x + \left\lbrack {n < M}\right\rbrack \cdot \left( {M \cdot p - n \cdot p}\ri... | academic_literature | english | double_column | false | false | 1,700 | 2,200 | |
page-28c45f5f-7e0d-464a-89ec-8de3a4abb927.png | Extract the text from the above document as if you were reading it naturally. Return the tables in html format. Return the equations in LaTeX representation. If there is an image in the document wrap it as follows <img>image here</img>. Watermarks should be wrapped in brackets. Ex: <watermark>OFFICIAL COPY</watermark>.... | J. Cardiovasc. Dev. Dis. 2025, 12, 13
<page_number>2 of 14</page_number>
recommendations of the Cardiological Organizational Committee for Sports Eligibility (COCIS) [1,2].
## Background
The prevalence of CAAs remains unclear and varies significantly depending on the diagnostic method used. It ranges from 1% to 5% ... | academic_literature | english | single_column | false | false | 1,654 | 2,339 | |
page-dca64e05-1ce6-49ee-857a-9b5d05a87357.png | Extract the text from the above document as if you were reading it naturally. Return the tables in html format. Return the equations in LaTeX representation. If there is an image in the document wrap it as follows <img>image here</img>. Watermarks should be wrapped in brackets. Ex: <watermark>OFFICIAL COPY</watermark>.... | <table><tr><td></td><td colspan="10">1</td><td colspan="10">2</td><td colspan="10">3</td><td colspan="10">4</td><td colspan="9"></td></tr><tr><td></td><td>1</td><td>2</td><td>3</td><td>4</td><td>5</td><td>6</td><td>7</td><td>8</td><td>9</td><td>0</td><td>1</td><td>2</td><td>3</td><td>4</td><td>5</td><td>6</td><td>7</td... | academic_literature | english | single_column | false | false | 1,653 | 2,339 | |
page-cb30bac7-d4bd-47d0-ac84-e0b46d051fe9.png | Extract the text from the above document as if you were reading it naturally. Return the tables in html format. Return the equations in LaTeX representation. If there is an image in the document wrap it as follows <img>image here</img>. Watermarks should be wrapped in brackets. Ex: <watermark>OFFICIAL COPY</watermark>.... | TABLE 4. Uncorrected p-distances (percentage) of mitochondrial H1 fragment (ca. 2400 bp) among species and lineages of Paratelmatobius. Intraspecific or intra-lineage distances are highlighted in gray (in diagonal). Interspecific or inter-lineage distances are shown under the diagonal. Data are shown as range (mean) wh... | academic_literature | english | single_column | false | false | 1,654 | 2,339 | |
page-d2cdac80-318c-4157-ac6e-bdab333f66bd.png | Extract the text from the above document as if you were reading it naturally. Return the tables in html format. Return the equations in LaTeX representation. If there is an image in the document wrap it as follows <img>image here</img>. Watermarks should be wrapped in brackets. Ex: <watermark>OFFICIAL COPY</watermark>.... | # Introduction
Assessment of subfamily classification in the Lampyridae (and the Luciolinae in particular) has undergone a transformation from a purely morphology-based system to phylogenetically-based features of males, females and larvae. Some analyses incorporate mitochondrial RNA and can be compared with more conv... | academic_literature | english | single_column | false | false | 1,653 | 2,339 | |
page-50ffe5d2-cd27-449a-a813-d4a8c44b12ea.png | Extract the text from the above document as if you were reading it naturally. Return the tables in html format. Return the equations in LaTeX representation. If there is an image in the document wrap it as follows <img>image here</img>. Watermarks should be wrapped in brackets. Ex: <watermark>OFFICIAL COPY</watermark>.... | ANALYSIS
NATURE ENERGY
<table><tr><td>Conversion pathway</td><td>Label</td><td>Description</td><td>Feedstock feasibility</td><td>Energy input</td><td>Energy output (main)</td><td>Energy output (coproducts)</td><td>Displaced products</td><td>Refs.</td></tr><tr><td>CHP</td><td>E1</td><td>Thermal combustion through biom... | academic_literature | english | single_column | false | false | 1,654 | 2,197 | |
page-5d58cdf7-6d00-4ea5-97b2-358cbc0cab21.png | Extract the text from the above document as if you were reading it naturally. Return the tables in html format. Return the equations in LaTeX representation. If there is an image in the document wrap it as follows <img>image here</img>. Watermarks should be wrapped in brackets. Ex: <watermark>OFFICIAL COPY</watermark>.... | R.H.A.J. Ogink et al.
Technovation 119 (2023) 102621
this respect, regions are meaningful loci of innovation, especially when they have developed into industrial clusters. Such regional innovation systems are fostered by direct and indirect linkages, cooperation and synergies among local economic actors and instituti... | academic_literature | english | other_layout | false | false | 1,654 | 2,205 |
OmniDocBench-EN-250
A compact, English, document-type-stratified OCR / document-parsing benchmark of 250 pages derived from OmniDocBench (CVPR 2025). Each sample is one page image paired with a ground-truth transcription formatted for a single fixed instruction prompt (tables as HTML, equations as LaTeX, images and page numbers wrapped in tags), so end-to-end vision-language OCR models can be evaluated with one prompt.
⚠️ All page images and annotations come from OmniDocBench by OpenDataLab. This repository only re-formats and sub-samples them. Please cite the original work (see Citation) and respect its license (CC BY-NC 4.0, research use only).
Quick start
from datasets import load_dataset
ds = load_dataset("Misraj/OmniDocBench-EN-250", split="test")
ex = ds[0]
ex["image"] # PIL.Image of the page
ex["prompt"] # the instruction below (identical for every row)
ex["text"] # ground-truth target
The prompt
Extract the text from the above document as if you were reading it naturally. Return the tables in html format. Return the equations in LaTeX representation. If there is an image in the document wrap it as follows <img>image here</img>. Watermarks should be wrapped in brackets. Ex: <watermark>OFFICIAL COPY</watermark>. Page numbers should be wrapped in brackets. Ex: <page_number>14</page_number>.
Data fields
| field | type | description |
|---|---|---|
image |
image | Page image from OmniDocBench |
image_name |
string | Original OmniDocBench image file name (traceable to the source) |
prompt |
string | The fixed instruction above |
text |
string | Ground-truth output formatted for the prompt |
data_source |
string | OmniDocBench document type (academic paper, book, slides, newspaper, …) |
language |
string | Page language from OmniDocBench page_attribute |
layout |
string | Page layout attribute (single column, double column, …) |
has_watermark |
bool | OmniDocBench page attribute watermark |
fuzzy_scan |
bool | OmniDocBench page attribute fuzzy_scan |
width, height |
int32 | Page image size in pixels |
How the dataset was built
Step 1 — Source and language filter
- Source:
opendatalab/OmniDocBench(1651 annotated pages, block-level layout, reading order and text / LaTeX / HTML annotations). - Kept pages whose
page_attribute.languageis English: 755 pages; 0 dropped because the page image was missing → 755 pages.
Step 2 — Converting annotations to the prompt format
The OmniDocBench layout_dets blocks of each page were sorted by their annotated
reading order (blocks without an order go last), rendered as below and joined
by a blank line. The target is plain text except for the tags/markup the
prompt asks for.
| OmniDocBench block | Rendered as |
|---|---|
header, footer |
plain text, kept in reading order |
text_block, title, figure_caption, figure_footnote, table_caption, table_footnote, equation_caption, page_footnote, reference, code_txt_caption |
plain text (no Markdown headings) |
equation_isolated |
$$ … $$ (original $, $$, \[ \], \( \) delimiters normalised) |
table |
HTML from the html annotation (<html>/<body> removed, whitespace between tags collapsed); LaTeX fallback if no HTML exists |
figure |
<img></img> (OmniDocBench has no figure descriptions, so the tag is empty) |
page_number |
<page_number>…</page_number> |
code_txt |
fenced code block |
abandon |
dropped (OmniDocBench noise regions) |
other blocks with ignore: true |
dropped (as in the official OmniDocBench evaluation); headers, footers and page numbers are always kept |
Notes:
- OmniDocBench does not annotate watermark text, so
<watermark>tags never appear in the ground truth. Pages with a watermark are flaggedhas_watermark = true. - Inline formulas stay inside their text block exactly as annotated by OmniDocBench.
Step 3 — Stratified sampling of 250 pages
250 of the 755 English pages were sampled, stratified by data_source:
each document type keeps the same share in the benchmark as in the full English set.
Quotas were computed as 250 × type_count / 755 and rounded with the
largest-remainder method so the total is exactly 250. Pages within each type
were sampled with random.Random(42).
Distribution by document type
| data_source | English set (n) | English set % | benchmark (n) | benchmark % |
|---|---|---|---|---|
| academic_literature | 206 | 27.3% | 68 | 27.2% |
| book | 131 | 17.4% | 43 | 17.2% |
| PPT2PDF | 120 | 15.9% | 40 | 16.0% |
| magazine | 84 | 11.1% | 28 | 11.2% |
| exam_paper | 82 | 10.9% | 27 | 10.8% |
| newspaper | 77 | 10.2% | 26 | 10.4% |
| colorful_textbook | 53 | 7.0% | 18 | 7.2% |
| research_report | 1 | 0.1% | 0 | 0.0% |
| note | 1 | 0.1% | 0 | 0.0% |
| total | 755 | 100% | 250 | 100% |
Document types too small to receive a benchmark page under proportional rounding: note (1 page(s)), research_report (1 page(s)).
Benchmark composition by layout
| layout | n | % |
|---|---|---|
| single_column | 109 | 43.6% |
| other_layout | 55 | 22.0% |
| double_column | 38 | 15.2% |
| 1andmore_column | 30 | 12.0% |
| three_column | 18 | 7.2% |
Pages with watermark: 0 · Fuzzy scans: 0 · Document types: 7
Suggested evaluation
- Run the model on
image+promptand compare its output withtext. - Recommended metrics (same family as OmniDocBench): normalised edit distance for text, TEDS for tables (HTML), CDM / edit distance for LaTeX formulas.
- OmniDocBench's official matching and metrics: evaluation code.
- Report scores per
data_sourceas well as overall.
Limitations
- English pages only.
- No watermark transcriptions; figures are empty
<img></img>tags. - 250 pages: good for fast iteration, but small document types contain only a few pages.
License
Inherited from OmniDocBench: CC BY-NC 4.0 — research and non-commercial use only. Original PDFs were collected by OpenDataLab from public online channels and community contributions; copyright questions about source pages should go to OpenDataLab (OpenDataLab@pjlab.org.cn).
Citation
If you use this benchmark, please cite OmniDocBench:
@inproceedings{ouyang2025omnidocbench,
title = {OmniDocBench: Benchmarking Diverse PDF Document Parsing with Comprehensive Annotations},
author = {Ouyang, Linke and Qu, Yuan and Zhou, Hongbin and Zhu, Jiawei and Zhang, Rui and
Lin, Qunshu and Wang, Bin and Zhao, Zhiyuan and Jiang, Man and Zhao, Xiaomeng and
Shi, Jin and Wu, Fan and Chu, Pei and Liu, Minghao and Li, Zhenxiang and Xu, Chao and
Zhang, Bo and Shi, Botian and Tu, Zhongying and He, Conghui},
booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
year = {2025},
note = {arXiv:2412.07626}
}
And optionally this benchmark:
@misc{misraj_omnidocbench_en_250,
title = {OmniDocBench-EN-250: an English, document-type-stratified OCR benchmark derived from OmniDocBench},
author = {Misraj},
year = {2026},
howpublished = {\url{https://huggingface.co/datasets/Misraj/OmniDocBench-EN-250}}
}
- Downloads last month
- 37