awesome-academic-phrase

Awesome Academic Phrase

A curated list of academic phrases in research papers.

Why Awesome Academic Phrase Collection?

Reading takes time. Along the way, we often come across well-crafted phrases worth keeping. This repository records and curates those expressions to:

• Learn lively: Observe how ideas are expressed in lively papers
• Reuse phrasing: Adapt strong expressions for your own writing
• Keep humanity: Preserve unexpected language usage in the age of AI

Contributing

Pull requests are welcome, please feel free to raise pull requests to add new recommendation.

Table of Contents

• General words
• Collocation
• Adjective
• Mathematical
• Algorithm
• Daily life
• Report
• Book
• Blog

General words

Expand to see all recorded general words...

- [conduct](https://dictionary.cambridge.org/dictionary/english/conduct): "The review will be **conducted** through the above broadened perspectives of human mobility." [(Wang et al., 2019)](https://doi.org/10.1016/j.cities.2022.103939) - [cater](https://dictionary.cambridge.org/dictionary/english/cater): "The present study proposes a universal approach to classifying intra-city tourists, which not only **caters** to the development of geo-big data but also incorporates the perspective of urban tourism functions." [(Park et al., 2023)](https://doi.org/10.1016/j.tourman.2022.104718) - [deviate](https://dictionary.cambridge.org/dictionary/english/deviate): "However, training LSTM with maximum likelihood estimation suffers from the exposure bias problem, where the generated samples might **deviate** from a realistic path with longer sequences." [(Kun et al., 2018)](https://doi.org/10.24963/ijcai.2018/530)

Collocation

• shed light on: “By examining travel flow interactions in transport systems, it is possible to shed light on the underlying structural characteristic of regions.” (Yang et al., 2020)

Adjective

• arbitrary: “The proposed approach builds on concepts from convolutional neural networks (CNNs) for images and extends them to arbitrary graphs.” (Niepert et al., 2016)

Mathematical

• general
  • quadratic: “While all graph kernels have a training complexity at least quadratic in the number of graphs, which is prohibitive for large-scale problems, PATCHY-SAN scales linearly with the number of graphs.” (Niepert et al., 2016)
  • spherical: “The spherical distance can not accurately estimate the expected cost.” (Jiang et al., 2023)
• graph
  • isomorphic: “The computational problem of deciding whether two graphs are isomorphic surfaces in several application domains.” (Niepert et al., 2016)

Algorithm

Expand to see all recorded algorithm words...

- [heuristic](https://dictionary.cambridge.org/dictionary/english/heuristic): "The A* algorithm is a **heuristic** search algorithm used extensively on the road network." [(Jiang et al., 2023)](https://doi.org/10.48550/arXiv.2301.07103) - [merit](https://dictionary.cambridge.org/dictionary/english/merit): "Both class of methods have their own **merits**, and transportation system applications require the right methods." [(Yu et al., 2019)](https://doi.org/10.1109/TITS.2019.2910560) - [vanilla](https://dictionary.cambridge.org/dictionary/english/vanilla): "To enhance the capability of LLMs in representing locations, we avoid using **vanilla** location IDs, as they lack inherent semantic information and fail to capture the contextual relationships between locations." [(Chen et al., 2025)](https://doi.org/10.1145/3711896.3736937)

Daily life

• thermal: thermal tumbler, thermal comfort

Report

A spectacular superconductor claim is making news. Here’s why experts are doubtful (ADRIAN CHO, 2023)

• “This week, social media has been aflutter over a claim for a new superconductor that works not only well above room temperatures, but also at ambient pressure. If true, the discovery would be one of the biggest ever in condensed matter physics and could usher in all sorts of technological marvels, such as levitating vehicles and perfectly efficient electrical grids. … On the other hand, he says, researchers at Argonne and elsewhere are already trying to replicate the experiment. … What’s more, the disorder introduced by the doping ought to further suppress superconductivity. … How will this be sorted out? The big question will be whether anybody can reproduce the observations. … “

^{[accessed on 1 Aug, 2023]}

ChatGPT-like AIs are coming to major science search engines (Richard Van Noorden, 2023)

Scopus, Dimensions and Web of Science are introducing conversational AI search.

• “… Many other AI search engine systems adopt a similar strategy, … Elsevier has also cut down the unpredictability of its AI by picking a low setting for the bot’s ‘temperature’ — a measure of how often it chooses to deviate from the most plausible words in its response. … Might users simply copy and paste the bot’s paragraphs into their own papers, effectively plagiarizing the tool? … Elsevier has so far tackled this with guidance that asks researchers to use the summaries responsibly, he says. Khan points out that funders and publishers have issued similar guidance, asking for transparent disclosure if LLMs are used in, … a search engine first retrieves relevant articles, … “

^{[accessed on 2 Aug, 2023]}

Is Fukushima wastewater release safe? What the science says (Bianca Nogrady, 2023)

Radiation in the water will be diluted to almost-background levels, but some researchers are not sure this will be sufficient to mitigate the risks

• “… Japan is pressing ahead with plans to release water contaminated by the 2011 meltdown of the Fukushima Daiichi nuclear power plant into the Pacific Ocean. … Jim Smith, …, says the risk this poses to nations around the Pacific Ocean will probably be negligible. … But Richmond is concerned the tritium could concentrate in the food web as larger organisms eat smaller contaminated ones. … Shigeyoshi Otosaka says that the organically bound form of tritium could accumulate in fish and marine organisms. … “We have confirmed that the tritium concentrations in the bodies of marine organisms reach equilibrium after a certain period of time and do not exceed the concentrations in the living environment,” the spokesperson said. … “

^{[accessed on 6 Sep, 2023]}

Three ways ChatGPT helps me in my academic writing (Dritjon Gruda, 2023)

Generative AI can be a valuable aid in writing, editing and peer review – if you use it responsibly, says Dritjon Gruda.

• “… The value that I derive from generative AI is not from the technology itself blindly churning out text, but from engaging with the tool and using my own expertise to refine what it produces. The dialogue between me and the chatbot both enhances the coherence of my work and, over time, teaches me how to describe complex topics in a simper way. … If something doesn’t quite hit the mark, don’t hesitate to say, “This isn’t quite what I meant. Let’s adjust this part.” Or you can commend its improvements: “This is much clearer, but let’s tweak the ending for a stronger transition to the next section.” … For instance, ChatGPT excels at explaining and justifying the reasons behind specific limitations that I had identified in my review, which helps me to grasp the broader implications of the study’s contribution. If I identify methodological limitations, ChatGPT can elaborate on these in detail and suggest ways to overcome them in a revision. … Occasionally, however, its suggestions are off-base, far-fetched, irrelevant or simply wrong. … ChatGPT has become indispensable in this process, helping me to craft precise, empathetic and actionable feedback without replacing human editorial decisions. … I’ve found that this approach both enhances the quality of my feedback and helps to guarantee that I convey my thoughts supportively. … These tools can bolster our capabilities in writing, reviewing and editing. They preserve the essence of scientific inquiry — curiosity, critical thinking and innovation — while improving how we communicate our research. … “

^{[accessed on 5 Jul, 2024]}

An open letter to graduate students and other procrastinators: it’s time to write (Dennis J. Hazelett, 2025)

• “… This letter is for the graduate student who has been asked by peers to begin writeup of a recent series of experiments. This person doesn’t known where to begin and has received or absorbed a lot of free and unsolicited advice — including what order to write in, how to arrange their office and schedule to maximize productivity, and perhaps some tips on self-care. … But you may approach the act of writing science with reluctance, as nuisance or distraction from the ‘real’, hands-on, day-to-day work. …, it may seem at times an insurmountable challenge to get started. … Quite the opposite: scientific writing is an iterative process in which ideas become more succinct and more cogent each time you pass over them. … All this really means is that there is no other way to tackle the largest tasks in life other than in the increment of work available to you, which may indeed be discouragingly small. But there is hope. … The second hard truth I have for you is this: thinking about your project in this way is actually a form of procrastination — quite possibly the most damaging kind. The only kind of thinking that matters in science is structured thinking. The only way to give structure and substance to your thoughts is to write them down. Writing. Is. Thinking. …, Get as far away from these people as possible. Anyone can write; it just takes practice. Regardless of your situation, you can still write for yourself, and your voice is great. … A rule of thumb is that your drafts should get shorter with each iteration. … Another reason why reading is hard, especially when it comes to technical papers, is that the vast majority of them are poorly written. … Sitting through a scientific talk is generally challenging, for the same reason that reading most papers is. They’re often prepared without much due given to the basic questions I posed at the beginning of this essay: what is it you mean to say, and to whom are you saying it? When you fail to communicate effectively, in any medium, you’ve wasted your time and that of others. Worse still, you’ve missed out on a precious opportunity to promote your work and that of your colleagues and mentors. The bottom line is: writing is the most essential activity of a scientist because without writing there is no thinking and no real opportunity for exchange in the marketplace of ideas. “

^{[accessed on 24 Apr, 2025]}

What would an AI university look like and how might it change education? (Jackson Ryan, 2025)

From lectures by avatars to entire qualifications, higher education centred around AI is just around the corner.

• “… AI has already had a seismic impact on higher education as existing universities race to integrate a suite of tools and LLMs into administration, curriculum design, teaching and assessment — and at the same time grapple with students’ use of LLMs. … Although the pitfalls associated with jamming AI into any place of learning are apparent, there is a need to equip students with the skills they will require beyond the education system. … He points to the vast disparities in the way that AI is being rolled out across institutions in South Africa, creating an imbalance in skills once students complete their studies. … “What we don’t yet have is an AI university that is truly intentional” in terms of its educational approach, he says. “One that says: this is our philosophy, this is our pedagogy, this is our ethical stance, and this is how AI fits into that vision. What we have instead is a rush. Everyone is hopping on the AI bandwagon and not wanting to be left behind.” “

^{[accessed on 6 Jan, 2026]}

Book

The Book of Why: The New Science of Cause and Effect (Judea Pearl and Dana Mackenzie, 2018)

• “Causal inference is all about taking this question seriously. It posits that the human brain is the most advanced tool ever devised for managing causes and effects. Our brains store an incredible amount of causal knowledge which, supplemented by data, we could harness to answer some of the most pressing questions of our time. More ambitiously, once we really understand the logic behind causal thinking, we could emulate it on modern computers and create an “artificial scientist.” This smart robot would discover yet unknown phenomena, find explanations to pending scientific dilemmas, design new experiments, and continually extract more causal knowledge from the environment.”

^{[recorded on 24 Nov, 2023]}

Handbook of Regional Science (Manfred M. Fischer and Peter Nijkamp, 2014)

Editorial Introduction

• “… It brings systematically together a varied set of … not purely …, but offer a tertiary literature review … The growth pace of regional science … the full spectrum of … provide a genuinue and appealing entry … location and agglomeration theory … the prominent historical perspective … Regional science from a locational angle did not only spur innovative, theoretical, and methodological research … In the same vein, housing markets and labor markets have become foci of regional science research … in a succinct way … a concise exposition … the cornerstones of regional science … Regional economic growth issues have inspired a wealth of research … the second section confirms largely the aforementioned sketch of regional economic growth issues … play a substantive role in … knowledge, networks, spillovers, and proximity … In the past decades, much attention has been devoted to … These findings are in agreement with the results of … This section on spatial connometrics offers an up-to-date overview of advances in the flourishing research domain. … Both figures depict largely the same type of information and may be seen as the main ingredients of this volume.”

^{[recorded on 17 Sep, 2024]}

Transitions Guide, APA Style 7th Edition (American Psychological Association, 2023)

To explain in a different way

• That is,
• In other words,

To illustrate

• Specifically,
• For example,
• As one example,
• In particular,

To emphasize a point/finding

• Interestingly,
• It is interesting that
• Surprisingly,
• It is surprising that
• Importantly,
• It is important that

Your own idea/argument

• It seems plausible that
• It seems reasonable that
• It seems logical that
• It could be argued that

To show cause/consequence

• As such,
• Consequently,
• Therefore,
• Thus,

To show contrast

• By contrast,
• Conversely,
• However,
• Although ___,
• Nevertheless,
• Nonetheless,
• ___ notwithstanding,
• On the one hand, … On the other hand,

To show additional examples

• Similarly,
• In a similar manner,
• In addition,
• Additionally,
• Moreover,
• Furthermore,
• Another study Note: besides sounds unplan, common used in speaking

To summarize/conclude

• In short,
• In sum,
• Overall,
• In general,
• Taken together, these findings

To show a sequence of events

• First,
• Next,
• Then,
• Finally

To relate to hypotheses

• Consistent with
• Contrary to
• As predicted,
• As expected,
• In agreement with
• In accordance with
• Congruent with

^{[recorded on 4 Dec, 2025]}

Blog

The spec is dead, long live the spec! (Ravi Mehta and Danny Martinez, 2025)

• “…Seen in that light, we’re doing AI development backwards. We craft careful prompts to communicate our intentions to models. The AI generates code. Then we keep the code and throw away the prompt. “This feels like you shred the source and then you very carefully version control the binary,” Grove observes. … William Gibson, the science fiction writer who coined the phrase “cyberspace,” once said: The future is already here—it’s just not very evenly distributed. I’m reminded of this when I think about what is and is not possible with AI. Today, AI’s gains are profoundly uneven. Some things—generating code, text, images—have made quantum leaps. They operate at AI speed. Other things—talking to customers, discovering their needs, convincing them to buy—still move at human speed.”

^{[accessed on 5 Mar, 2026]}

How to Fix Your Context (Drew Breunig, 2025)

• “…When prompting DeepSeek-v3, the team found that selecting the the right tools becomes critical when you have more than 30 tools. Above 30, the descriptions of the tools begin to overlap, creating confusion. Beyond 100 tools, the model was virtually guaranteed to fail their test. Using RAG techniques to select less than 30 tools yielded dramatically shorter prompts and resulted in as much as 3x better tool selection accuracy (Gan and Sun, 2025).
  For smaller models, the problems begin long before we hit 30 tools. One paper we touched on last post, Less is More,” demonstrated that Llama 3.1 8b fails a benchmark when given 46 tools, but succeeds when given only 19 tools (Paramanayakam et al., 2024). The issue is context confusion, not context window limitaions.”

^{[accessed on 6 Mar, 2026]}

Peeking Under the Hood of Claude Code (OutSight AI, 2025)

• “…(TL;DR) Claude Code front-loads context with tiny, targeted prompts (titles, topic checks, summaries) before doing real work. It sprinkles “system-reminders” everywhere including system/user prompts, tool calls, even tool results, to reduce drift.”

Context Rot: How Increasing Input Tokens Impacts LLM Performance (Chroma Research, 2025)

• “Large Language Models (LLMs) are typically presumed to process context uniformly—that is, the model should handle the 10,000th token just as reliably as the 100th. However, in practice, this assumption does not hold. We observe that model performance varies significantly as input length changes, even on simple tasks. In this report, we evaluate 18 LLMs, including the state-of-the-art GPT-4.1, Claude 4, Gemini 2.5, and Qwen3 models. Our results reveal that models do not use their context uniformly; instead, their performance grows increasingly unreliable as input length grows.”

^{[accessed on 7 Mar, 2026]}

Humanity’s Last Gasp (Latent.Space, 15 Apr 2026)

Aaron Levie reports that “AI is not causing anyone to do less work right now, and similar to Silicon Valley people feel their teams are the busiest they’ve ever been.”

• “How can it both be true that “Agents are doing more work and yet Everyone is working harder”? How can it be true that Claude Mythos has been used internally for 2 months, and yet Claude keeps going down? How can it be true that Model and Agent Labs are more productive than ever and yet acquihiring and acquiring more than ever? A simple thought exercise we’ve made before is the “Turkey problem”, where, based on real evidence and an abundance of historical data, Turkeys should conclude that life is fantastic and all of humanity is set up to make turkeys well fed as far as they’ve ever experienced. Turkey doomsayers would be alarmist, crackpots, and then ignored. Until Thanksgiving. Are engineers, or all knowledge workers in general, turkeys, in this scenario? Should our “elasticity” and value of work be increasingly positive, right up to some crossover point we become horses? Now that SWE-Bench is saturated (with SWE-Bench Pro soon to be, Mythos is at 78%) and GDPval rates GPT 5.4 as better than/equal to human experts 83% of the time in most swathes of the economy, what’s left?”

^{[accessed on 18 Apr, 2026]}