Delete tokenized

tokenized/biology_test.bin

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:d1bbb46de328a5fb15d012bcda25835f4a4984380929323106a208fe00260da2
-size 130256772

tokenized/biology_train.bin

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:ff821dc52b2b83228cae6813232e4bd0e0eedb1d22c59a88e8d4cf81d5e0b4bc
-size 1126247082

tokenized/cyber_test.bin

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:ff7ada52c7442ee3e0adadce2c1203e0dbc1dba993ffc5870256e8645253ced7
-size 188492922

tokenized/cyber_train.bin

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:d15c0130aa1fc46000fbe9adfdfbcbf28ff4f2433615bd2d579d2d5a3ac51ea6
-size 1685193724

tokenized/metadata.json

@@ -1,77 +0,0 @@
-{
-  "biology": {
-    "train": {
-      "file": "biology_train.bin",
-      "num_papers": 27721,
-      "num_tokens": 563123541,
-      "avg_tokens_per_paper": 20313.96922910429,
-      "max_tokens": 363998,
-      "min_tokens": 14,
-      "example_text_preview": "# \\[1612.09478\\] Discovery of cancer common and specific driver gene sets\n\n# Discovery of cancer common and specific driver gene sets\n\nJunhua Zhang and Shihua Zhang J. Zhang and S. Zhang are with National Center for Mathematics and Interdisciplinary Sciences, Academy of Mathematics and Systems Science, Chinese Academy of Sciences, Beijing 100190, China.\nE-mail: [email protected], [email protected].\n\n###### Abstract\n\nCancer is known as a disease mainly caused by gene alterations. Discovery of mutated driver pathways or gene sets is becoming an important step to understand molecular mechanisms of carcinogenesis. However, systematically investigating commonalities and specificities of driver gene sets among multiple cancer types is still a great challenge, but this investigation will undoubtedly benefit deciphering cancers and will be helpful for personalized therapy and precision medicine in cancer treatment. In this study, we propose two optimization models to *de novo* discover common driver gene sets among multiple cancer types (ComMDP) and specific driver gene sets of one certain or multiple cancer types to other cancers (SpeMDP), respectively. We first apply ComMDP and SpeMDP to simulated data to validate their efficiency. Then, we further apply these methods to 12 cancer types from The Cancer Genome Atlas (TCGA) and obtain several biologically meaningful driver pathways. As examples, we construct a common cancer pathway model for BRCA and OV, infer a complex driver pathway model for BRCA carcinogenesis based on common driver gene sets of BRCA with eight cancer types, and investigate specific driver pathways of the liquid cancer lymphoblastic acute myeloid leukemia (LAML) versus other solid cancer types. In these processes more candidate cancer genes are also found.\n\n###### Index Terms:\n\nBioinformatics, cancer genomics, pan-cancer study, driver pathway, mutual exclusivity\n\n## 1 Introduction\n\nCancer is a complex and heterogeneous disease with diverse genetic and environmental factors involved in its etiology. With the advances of deep sequencing technology, huge volume cancer genomics data have been generated through several large-scale programs (e.g., The Cancer Genome Atlas (TCGA) \\[[1](#bib.bib1)\\], International Cancer Genome Consortium (ICGC) \\[[2](#bib"
-    },
-    "test": {
-      "file": "biology_test.bin",
-      "num_papers": 3081,
-      "num_tokens": 65128386,
-      "avg_tokens_per_paper": 21138.71665043817,
-      "max_tokens": 237014,
-      "min_tokens": 39,
-      "example_text_preview": "# \\[2503.02351\\] MindSimulator: Exploring Brain Concept Localization via Synthetic FMRI\n\n# MindSimulator: Exploring Brain Concept Localization via Synthetic FMRI\n\nGuangyin Bao, Qi Zhang, Zixuan Gong, Zhuojia Wu, Duoqian Miao\n\nGuangyin Bao<sup>1</sup>, Qi Zhang<sup>1</sup>, Zixuan Gong<sup>1</sup>, Zhuojia Wu<sup>1</sup>, Duoqian Miao<sup>1</sup>\n<sup>1</sup>Tongji University\nCorresponding author.\n\n###### Abstract\n\nConcept-selective regions within the human cerebral cortex exhibit significant activation in response to specific visual stimuli associated with particular concepts. Precisely localizing these regions stands as a crucial long-term goal in neuroscience to grasp essential brain functions and mechanisms. Conventional experiment-driven approaches hinge on manually constructed visual stimulus collections and corresponding brain activity recordings, constraining the support and coverage of concept localization. Additionally, these stimuli often consist of concept objects in unnatural contexts and are potentially biased by subjective preferences, thus prompting concerns about the validity and generalizability of the identified regions. To address these limitations, we propose a data-driven exploration approach. By synthesizing extensive brain activity recordings, we statistically localize various concept-selective regions. Our proposed MindSimulator leverages advanced generative technologies to learn the probability distribution of brain activity conditioned on concept-oriented visual stimuli. This enables the creation of simulated brain recordings that reflect real neural response patterns. Using the synthetic recordings, we successfully localize several well-studied concept-selective regions and validate them against empirical findings, achieving promising prediction accuracy. The feasibility opens avenues for exploring novel concept-selective regions and provides prior hypotheses for future neuroscience research.\n\n## 1 Introduction\n\nThe human brain’s visual cortex is decisive in processing and perceiving visual information. Neuroscience researchers have long dedicated themselves to unraveling the brain’s visual mechanisms, making impressive strides such as in brain visual encoding (Mitchell et al., [2008](#bib.bib35)), decoding (Gong et al., [2024b](#bib.bib14)), and visual perception (Chen et al., [2020]("
-    }
-  },
-  "cyber": {
-    "train": {
-      "file": "cyber_train.bin",
-      "num_papers": 33023,
-      "num_tokens": 842596862,
-      "avg_tokens_per_paper": 25515.454743663508,
-      "max_tokens": 577123,
-      "min_tokens": 1,
-      "example_text_preview": "# Intractable Group-theoretic Problems Around Zero-knowledge Proofs\n\nCansu Betin Onur\nInstitute of Applied Mathematics\nMiddle East Technical University\n06800 Ankara Turkey EMail: [email protected]\n\n###### Abstract\n\nWhile the amount of data produced and accumulated continues to advance at unprecedented rates, protection and concealment of data increase its prominence as a field of scientific study that requires more action. It is essential to protect privacy-sensitive data at every phase; at rest, at run, and while computations are executed on data. The zero-knowledge proof (ZKP) schemes are a cryptographic tool toward this aim. ZKP allows a party to securely ensure the data’s authenticity and precision without revealing confidential or privacy-sensitive information during communication or computation. The power of zero-knowledge protocols is based on intractable problems. There is a requirement to design more secure and efficient zero-knowledge proofs. This demand raises the necessity of determining appropriate intractable problems to develop novel ZKP schemes. In this paper, we present a brief outline of ZKP schemes, the connection of these structures to group-theoretic intractable problems, and annotate a list of intractable problems in group theory that can be employed to devise new ZKP schemes.\n\n###### Contents\n\n1.  [](#S1 \"In Intractable Group-theoretic Problems Around Zero-knowledge Proofs\")\n\n    1 Introduction\n\n2.  [](#S2 \"In Intractable Group-theoretic Problems Around Zero-knowledge Proofs\")\n\n    2 Zero Knowledge Proof Systems\n\n    1.  [](#S2.SS1 \"In 2 Zero Knowledge Proof Systems ‣ Intractable Group-theoretic Problems Around Zero-knowledge Proofs\")\n\n        2.1 Basic Concept\n\n    2.  [](#S2.SS2 \"In 2 Zero Knowledge Proof Systems ‣ Intractable Group-theoretic Problems Around Zero-knowledge Proofs\")\n\n        2.2 A Brief History of ZKP\n\n    3.  [](#S2.SS3 \"In 2 Zero Knowledge Proof Systems ‣ Intractable Group-theoretic Problems"
-    },
-    "test": {
-      "file": "cyber_test.bin",
-      "num_papers": 3670,
-      "num_tokens": 94246461,
-      "avg_tokens_per_paper": 25680.234604904632,
-      "max_tokens": 270429,
-      "min_tokens": 1,
-      "example_text_preview": "# Cutting through buggy adversarial example defenses: fixing 1 line of code breaks Sabre\n\nNicholas Carlini\n*Google DeepMind*\n\n###### Abstract\n\nSabre is a defense to adversarial examples that was accepted at IEEE S&P 2024. We first reveal significant flaws in the evaluation that point to clear signs of gradient masking. We then show the cause of this gradient masking: a bug in the original evaluation code. By fixing a single line of code in the original repository, we reduce Sabre’s robust accuracy to 0%. In response to this, the authors modify the defense and introduce a new defense component not described in the original paper. But this fix contains a second bug; modifying one more line of code reduces robust accuracy to *below* baseline levels. After we released the first version of our paper online, the authors introduced another change to the defense; by commenting out one line of code during attack we reduce the robust accuracy to 0% again.\n\n## 1 Introduction\n\nSabre \\[[5](#bib.bib5)\\] is a defense to adversarial examples that claims to be $`3 \\times`$ more robust to attack than the current state-of-the-art \\[robustbench\\], bringing the best attacks on CIFAR-10 at a distortion of $`\\varepsilon = {8/255}`$ from 29% success rate down to just 11%. In this paper we begin with a critique of the Sabre paper, and show it contains multiple evaluation flaws. For example, the paper claims that Sabre is more accurate when an adversary is attacking the defense than when not under attack, claims mathematically impossible levels of robustness, and does not follow standard recommended best practices \\[[3](#bib.bib3)\\] when evaluating adversarial robustness—including omitting any analysis against adaptive attacks \\[[7](#bib.bib7)\\].\n\nNevertheless, Sabre was accepted at IEEE S&P 2024, one of the top computer security conferences. (Readers may be forgiven for mistaking this defense for a *different* defense—also with a flawed evaluation—accepted at last year’s IEEE S&P 2023. \\[[2](#bib.bib2)\\])\n\nIn this short paper we show that by changing a single line of code to"
-    }
-  },
-  "nuclear": {
-    "train": {
-      "file": "nuclear_train.bin",
-      "num_papers": 40491,
-      "num_tokens": 1004608274,
-      "avg_tokens_per_paper": 24810.656047022796,
-      "max_tokens": 1315605,
-      "min_tokens": 1,
-      "example_text_preview": "# Dissipative hydrodynamic evolution of hot quark matter at finite baryon density\n\nDepartment of Physics, The University of Tokyo, Tokyo 113-0033, Japan\nTheoretical Research Division, Nishina Center, RIKEN, Wako 351-0198, Japan\nE-mail\n\n###### Abstract:\n\nHigh-energy heavy ion collider experiments at RHIC and LHC have revealed that relativistic hydrodynamic models describe the hot and dense quark matter quantitatively. In this study, I develop a novel dissipative hydrodynamic model at finite baryon density to investigate the net baryon rapidity distribution. The results show that the distribution is widened in hydrodynamic evolution, which implies that the transparency of the collisions is effectively enhanced. This suggests that the kinetic energy loss for medium production at the initial stage could be larger. Furthermore, the net baryon distribution is found sensitive to baryon diffusion, implying that dissipative hydrodynamic modeling would be important for understanding the hot medium.\n\n## 1 Introduction\n\nQuarks and gluons are considered to be deconfined from hadrons at high temperatures and to form the quark-gluon plasma (QGP) \\[[1](#bib.bib1)\\]. It had been naïvely expected that the plasma is weakly coupled because of the asymptotic freedom of QCD. High-energy heavy ion collisions revealed, on the contrary, that the hot QCD matter would be strongly coupled near the quark-hadron crossover. The experimentally observed hadronic yields turned out to be described quantitatively by relativistic hydrodynamic models with small viscosity at the Relativistic Heavy Ion Collider (RHIC) and the Large Hadron Collider (LHC). Modern hydrodynamic analyses take account of viscosity and fluctuation to improve the model accuracy \\[[2](#bib.bib2)\\]. However, the net baryon number is neglected in most of those studies even though it would be important in understanding the experimental results such as particle-to-antiparticle ratios. This is possibly because viscous hydrodynamic equations of motion are complicated in the presence of finite charge density and also the boost invariant ansatz is often assumed, by which the net baryon number conserved at forward rapidity is omitted.\n\nBaryon"
-    },
-    "test": {
-      "file": "nuclear_test.bin",
-      "num_papers": 4499,
-      "num_tokens": 112801686,
-      "avg_tokens_per_paper": 25072.613025116694,
-      "max_tokens": 418966,
-      "min_tokens": 857,
-      "example_text_preview": "# \\[nucl-th/0008035\\] COMPACT AND LOOSELY BOUND STRUCTURES IN LIGHT NUCLEI\n\n# COMPACT AND LOOSELY BOUND STRUCTURES IN LIGHT NUCLEI\n\nO.L.Savchenko\n\nA.I.Steshenko\n\nBogolyubov Institute for Theoretical Physics of the NAS of Ukraine\nMetrolohichna Str., 14b, Kyiv-143, 03143 Ukraine\n\n###### Abstract\n\nA role of different components in the wave function of the weakly bound light nuclei states was studied within the framework of the cluster model, taking into account of orbitals ”polarization”. It was shown that a limited number of structures associated with the different modes of nucleon motion can be of great importance for such systems. Examples of simple and quite flexible trial wave functions are given for the nuclei <sup>8</sup>Be, <sup>6</sup>He. Expressions for the microscopic wave functions of these nuclei were found and used for the calculation of basic nuclear characteristics, using well known central-exchange nucleon-nucleon potentials.\n\n###### pacs:\n\n64.60.Cn, 81.30.Bx, 75.10.H\n\n## I Introduction.\n\nTheoretical consideration of strongly interacting many-particle systems, such as the atomic nuclei, has a number of difficulties, one of which is constructing of trial wave function. The wave function used should be characterized by proper set of the definite quantum numbers and corresponding symmetries, taking into account the most important physical modes of nucleon motion. It is usually assumed that structure of the constructed wave function is well known in the limitsd cases, i.e. in the case where a distance between fragments is very large or when we have a compact one-centre system. Hence, the main attention is paid to determination of correct interpolation expression for the wave function in the intermediate region.\n\nMany authors have worked at simultaneous description of both ”compact” and ”friable” structures (see, for example, Ref.\\[[1](#bib.bib1), [2](#bib.bib2)\\]). Several decades ago due to activity of Yu.A.Simon"
-    }
-  },
-  "all": {
-    "total_papers": 112485,
-    "total_tokens": 2682505210,
-    "categories": [
-      "biology",
-      "cyber",
-      "nuclear"
-    ],
-    "vocab_size": 50257,
-    "tokenizer": "EleutherAI/gpt-neo-125M",
-    "max_length": null,
-    "dtype": "uint16",
-    "train_test_split": "90/10",
-    "shuffle_seed": 42
-  }
-}

tokenized/nuclear_test.bin

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:517bcb57bd659ca1a88ba9e1378475949aff7f36b308079b3c88c3a1d0c41b5e
-size 225603372

tokenized/nuclear_train.bin

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:b4e501596daea09082c7adf05d9d9751bb673ef228f1232fa13cdd5523c42c56
-size 2009216548