metadata

datasets:
  - tum-nlp/IDMGSP
language:
  - en
tags:
  - scientific paper
  - fake papers
  - science
  - scientific text
widget:
  - text: >
      Abstract:


      The Hartree-Fock (HF) method is a widely used method for approximating the
      electronic structure of many-electron systems. In this work, we study the
      properties of HF solutions of the three-dimensional electron gas (3DEG), a
      model system consisting of a uniform, non-interacting electron gas in
      three dimensions. We find that the HF solutions accurately reproduce the
      known analytic results for the ground state energy and the static
      structure factor of the 3DEG. However, we also find that the HF solutions
      fail to accurately describe the excitation spectrum of the 3DEG,
      particularly at high energies.


      Introduction:


      The HF method is a self-consistent method for approximating the electronic
      structure of many-electron systems. It is based on the assumption that the
      electrons in a system can be described as non-interacting quasiparticles,
      each with its own effective potential. The HF method is commonly used to
      study the ground state properties of systems, such as the energy and the
      density distribution, but it can also be used to study excited states.


      The 3DEG is a model system that has been widely studied as a test case for
      electronic structure methods. It consists of a uniform, non-interacting
      electron gas in three dimensions, with a finite density and a periodic
      boundary condition. The 3DEG has a number of known analytic results for
      its ground state properties, such as the ground state energy and the
      static structure factor, which can be used to test the accuracy of
      approximate methods.


      Conclusion:


      In this work, we have studied the properties of HF solutions of the 3DEG.
      We find that the HF solutions accurately reproduce the known analytic
      results for the ground state energy and the static structure factor of the
      3DEG. However, we also find that the HF solutions fail to accurately
      describe the excitation spectrum of the 3DEG, particularly at high
      energies. This suggests that the HF method may not be suitable for
      accurately describing the excited states of the 3DEG. Further work is
      needed to understand the limitations of the HF method and to develop
      improved methods for studying the electronic structure of many-electron
      systems.
    example_title: Example ChatGPT fake
  - text: >
      Abstract:


      Recent calculations have pointed to a 2.8 $\sigma$ tension between data on
      $\epsilon^{\prime}_K / \epsilon_K$ and the standard-model (SM) prediction.
      Several new physics (NP) models can explain this discrepancy, and such NP
      models are likely to predict deviations of $\mathcal{B}(K\to \pi \nu
      \overline{\nu})$ from the SM predictions, which can be probed precisely in
      the near future by NA62 and KOTO experiments. We present correlations
      between $\epsilon^{\prime}_K / \epsilon_K$ and $\mathcal{B}(K\to \pi \nu
      \overline{\nu})$ in two types of NP scenarios: a box dominated scenario
      and a $Z$-penguin dominated one. It is shown that different correlations
      are predicted and the future precision measurements of $K \to \pi \nu
      \overline{\nu}$ can distinguish both scenarios.


      Introduction:


      CP violating ﬂavor-changing neutral current decays of K mesons are
      extremely sensitive to new physics (NP) and can probe virtual eﬀects of
      particles with masses far above the reach of the Large Hadron Collider.
      Prime examples of such observables are ϵ′ K measuring direct CP violation
      in K → ππ decays and B(KL → π0νν). Until recently, large theoretical
      uncertainties precluded reliable predictions for ϵ′ K. Although
      standard-model (SM) predictions of ϵ′ K using chiral perturbation theory
      are consistent with the experimental value, their theoretical
      uncertainties are large. In contrast, calculation by the dual QCD approach
      1 ﬁnds the SM value much below the experimental one. A major breakthrough
      has been the recent lattice-QCD calculation of the hadronic matrix
      elements by RBC-UKQCD collaboration 2, which gives support to the latter
      result. The SM value at the next-to-leading order divided by the indirect
      CP violating measure ϵK is 3 which is consistent with (ϵ′ K/ϵK)SM =
      (1.9±4.5)×10−4 given by Buras et al 4.a Both results are based on the
      lattice numbers, and further use CP-conserving K → ππ data to constrain
      some of the hadronic matrix elements involved. Compared to the world
      average of the experimental results 6, Re (ϵ′ K/ϵK)exp = (16.6 ± 2.3) ×
      10−4, (2) the SM prediction lies below the experimental value by 2.8 σ.
      Several NP models including supersymmetry (SUSY) can explain this
      discrepancy. It is known that such NP models are likely to predict
      deviations of the kaon rare decay branching ratios from the SM
      predictions, especially B(K → πνν) which can be probed precisely in the
      near future by NA62 and KOTO experiments.b In this contribution, we
      present correlations between ϵ′ K/ϵK and B(K → πνν) in two types of NP
      scenarios: a box dominated scenario and a Z-penguin dominated one.
      Presented at the 52th Rencontres de Moriond electroweak interactions and
      uniﬁed theories, La Thuile, Italy, 18-25 March, 2017. aOther estimations
      of the SM value are listed in Kitahara et al 5. b The correlations between
      ϵ′ K/ϵK, B(K → πνν) and ϵK through the CKM components in the SM are
      discussed in Ref. 7.


      Conclusion:


      We have presented the correlations between ϵ′ K/ϵK, B(KL → π0νν), and B(K+
      → π+νν) in the box dominated scenario and the Z-penguin dominated one. It
      is shown that the constraint from ϵK produces diﬀerent correlations
      between two NP scenarios. In the future, measurements of B(K → πνν) will
      be signiﬁcantly improved. The NA62 experiment at CERN measuring B(K+ →
      π+νν) is aiming to reach a precision of 10 % compared to the SM value
      already in 2018. In order to achieve 5% accuracy more time is needed.
      Concerning KL → π0νν, the KOTO experiment at J-PARC aims in a ﬁrst step at
      measuring B(KL → π0νν) around the SM sensitivity. Furthermore, the
      KOTO-step2 experiment will aim at 100 events for the SM branching ratio,
      implying a precision of 10 % of this measurement. Therefore, we conclude
      that when the ϵ′ K/ϵK discrepancy is explained by the NP contribution,
      NA62 experiment could probe whether a modiﬁed Z-coupling scenario is
      realized or not, and KOTO-step2 experiment can distinguish the box
      dominated scenario and the simpliﬁed modiﬁed Z-coupling scenario.
    example_title: Example real
license: openrail++

Model Card for IDMGSP-Galactica-TRAIN+GPT3

A fine-tuned Galactica model to detect machine-generated scientific papers based on their abstract, introduction, and conclusion.

This model is trained on the train+gpt3 dataset found in https://huggingface.co/datasets/tum-nlp/IDMGSP.

this model card is WIP, please check the repository, the dataset card and the paper for more details.

Model Details

Model Description

Developed by: Technical University of Munich (TUM)
Model type: [More Information Needed]
Language(s) (NLP): English
License: [More Information Needed]
Finetuned from model [optional]: Galactica

Model Sources

Repository: https://github.com/qwenzo/-IDMGSP
Paper: [More Information Needed]

Uses

Direct Use

from transformers import AutoTokenizer, OPTForSequenceClassification, pipeline

model = OPTForSequenceClassification.from_pretrained("tum-nlp/IDMGSP-Galactica-TRAIN+GPT3")
tokenizer = AutoTokenizer.from_pretrained("tum-nlp/IDMGSP-Galactica-TRAIN+GPT3")
reader = pipeline("text-classification", model=model, tokenizer = tokenizer)
reader(
'''
Abstract:
....

Introduction:
....

Conclusion:
...'''
)

Downstream Use [optional]

[More Information Needed]

Out-of-Scope Use

[More Information Needed]

Bias, Risks, and Limitations

[More Information Needed]

Training Details

Training Data

The training dataset comprises scientific papers generated by the Galactica, GPT-2, and SCIgen models, as well as papers extracted from the arXiv database.

The provided table displays the sample counts from each source utilized in constructing the training dataset. The dataset could be found in https://huggingface.co/datasets/tum-nlp/IDMGSP.

Dataset	arXiv (real)	ChatGPT (fake)	GPT-2 (fake)	SCIgen (fake)	Galactica (fake)	GPT-3 (fake)
TRAIN plus GPT-3 (TRAIN+GPT3)	8k	2k	2k	2k	2k	1.2k

Training Procedure

Preprocessing [optional]

[More Information Needed]

Training Hyperparameters

[More Information Needed]

Speeds, Sizes, Times [optional]

[More Information Needed]

Evaluation

Testing Data, Factors & Metrics

Testing Data

[More Information Needed]

Factors

[More Information Needed]

Metrics

[More Information Needed]

Results

[More Information Needed]

Summary

Model Examination [optional]

[More Information Needed]

Environmental Impact

Hardware Type: [More Information Needed]
Hours used: [More Information Needed]
Cloud Provider: [More Information Needed]
Compute Region: [More Information Needed]
Carbon Emitted: [More Information Needed]

Technical Specifications [optional]

Model Architecture and Objective

[More Information Needed]

Compute Infrastructure

[More Information Needed]

Hardware

[More Information Needed]

Software

[More Information Needed]

Citation [optional]

BibTeX:

[More Information Needed]

APA:

[More Information Needed]

Glossary [optional]

[More Information Needed]

More Information [optional]

[More Information Needed]

Model Card Authors [optional]

[More Information Needed]

Model Card Contact

[More Information Needed]