Collection of great vocabulary for paper writing
1 on the fly
dynamically, while in motion or progress
example: During training, the two components incorporate the estimated depth to produce supervisory signals on the fly. (copy from StructDepth)
2 data starvation/ supervision starvationre
3 hinting at a possibility for a hint to the possibility, provide a potential for
example: In contrast, recent large language models exhibit a wide range of capabilities, hinting at a possibility for similarly versatile models in computer vision. (copy from 4M)
4 out of box immediate usability (typically an electronic device or a piece of software), unusually good (the novel is nothing out of the box).
example: 4M leads to models that exhibit several key capabilities: (1) they can perform a diverse set of vision tasks out of the box. (copy from 4M)
5 remarkable flexibility can be used for describing controllability or editing,
example: enabling a wide variety of expressive multimodal editing capabilities with remarkable flexibility. (copy from 4M)
6 The field has seen a shift towards for the first paragraph in Introduction, the beginning of a storyline.
example: In recent years, the field of natural language processing (NLP) has seen a shift toward training large language models (LLMs) that are inherently capable of performing a wide range of tasks without requiring extensive task-specific adaptations [12, 25]. (copy from 4M)
7 there remains a need to Translation in Introduction. Translation from previous works to problems/needs or our work.
example: While these models have demonstrated remarkable success in NLP, there remains a need to develop similarly versatile and scalable models for vision.. (copy from 4M)
8 delve into to try hard to find out more information about something.
In this report, we delve into the performance of LLMs within the context of scientific discovery/research, focusing on GPT-4 (copy from Imapct of LLM on Scientific Discovery).
9 Broadly speaking without regard to specific details or exceptions
example: Broadly speaking, we evaluate GPT-4’s knowledge base, scientific understanding, scientific numerical calculation abilities, and various scientific prediction capabilities (copy from Imapct of LLM on Scientific Discovery).
10 transform the way
examples: i)Transform The Way You Work, and ii) LLMs are capable of transforming the way we generate and process information (copy from Imapct of LLM on Scientific Discovery).
11 transform the way
examples: i)Transform The Way You Work, and ii) LLMs are capable of transforming the way we generate and process information (copy from Imapct of LLM on Scientific Discovery).
12 extraordinary capabilities
example: Because of its extraordinary capabilities in general AI tasks, GPT-4 is also garnering significant attention in the scientific community.
13 central focus
example: Our aim is to provide a broad overview of LLMs’ performance and their potential applicability in these specific scientific fields, with GPT-4, the state-of-the-art LLM, as our central focus.
14 and beyond
example: These strive to uncover the fundamental principles and laws governing the universe, spanning from the smallest subatomic particles to the largest galaxies and beyond.
15 a wide array of
example: Natural science is an incredibly diverse field, encompassing a wide array of disciplines, including both physical sciences, which focus on non-living systems, and life sciences, which investigate living organisms.
16 overlaps with a part of the first thing occupies the same area as a part of the other thing.
example: It is important to note that these areas are not mutually exclusive; for example, drug discovery substantially overlaps with biology.
17 by which By which = How or whereby or “through which”
example: Drug discovery is the process by which new candidate medications are identified and developed to treat or prevent specific diseases and medical conditions.
18 on most fronts of most aspects
example: outperform these methods on most fronts (copy from 3D Gaussian splatting).
18 follow-up methods
example: The success of NeRF has resulted in an explosion of follow-up methods that address quality and speed, often by introducing regularization strategies. (copy from 3D Gaussian splatting).
19 equal
example: We are able to equal or in some cases surpass this quality while providing fast training and real-time rendering. (copy from 3D Gaussian splatting).
20 pave the way for the application of xx in
example:
21 charting future research paths
a multitude of studies have surfaced with the aim of charting future research paths, which have varied from identifying emerging trends to highlighting areas poised for swift progress.
22 despite the curse of
There is problem …… High-quality samples generated with score-based reverse diffusion algorithms provide evidence that deep neural networks (DNN) trained for denoising can learn high-dimensional densities, despite the curse of dimensionality.
23 ever-more
Deep neural networks (DNNs) have demonstrated ever-more impressive capabilities for learning and sampling from high-dimensional image densities.
作者：lingjivoo，github主页：传送门