Reading Papers, No Need to Try Too Hard

I’ve already forgotten when I developed the habit of reading papers.

But it definitely wasn’t during my master’s studies.

Beginner​

I still remember when I was first becoming a master’s student, feeling the endless authority of my advisor and listening to the senior students’ repeated advice:

"Reading papers is easy. Just read one every breakfast, and you can finish seven papers in a week. Writing a master’s thesis only takes three months!"

I almost spat my coffee on the computer when I heard that.

The story after that was nothing extraordinary. It took me two years to complete my master’s degree, and I barely read any papers.

Even though I later found out that what the seniors said was true, unfortunately, I was still too inexperienced at that time.

Research and Development Engineer​

I truly began reading papers after leaving the academic system.

At that point, no one was forcing me to read, no more presentation checks, and no more midterms or finals.

All I had were models that wouldn’t work, losses that wouldn’t converge, and architectures where the accuracy was always stuck at 72%.

I started searching for information, looking up keywords, opening paper after paper that I couldn’t understand.

Why should others be able to do it and not me?

Over time, those fragmented readings began to stack up, forming an unspoken sense of perception:

• Which model ideas might work
• Which architectures have reached their limit
• Which diagrams are deliberately made to look pretty
• Which experimental designs are actually meaningful

It wasn’t some kind of epiphany, but rather just practice that anyone could do.

Even now, I’m not someone who reads seven papers every day.

Nor have I read every paper in its entirety.

But I slowly found my own way of reading:

• Read up to 60 to 80%.

Not too casual, but not obsessive either; it’s like opening a box and seeing if there are any useful materials inside.

Reading Papers, No Need to Try Too Hard​

If you’re also on this path, I’d like to share a bit of advice: You don’t need to try too hard when reading papers.

You don’t get anxious when reading novels, so why are you anxious when reading papers?

For most papers, reading up to 60% is enough. You only need to understand the context of the article and find useful ideas and materials.

If you want to read more, you can try following these guidelines:

• 70%: Understand the model design and training strategy well enough to explain it to others.
• 80%: Reproduce the experiment and understand the boundaries between its success and failure.
• 90%: Modify, extend, or overturn it, making it part of your methodology.

Life is short, and you don’t have that much time to reach 90%. You need to learn how to start acting when you’re at 60%.

This doesn’t mean being sloppy; it means acknowledging the limited resources in reality.

Reading papers while implementing and making mistakes may seem slow, but it’s alive.

You’ll realize that many ideas only need 60% to be useful.

You can:

• Try a simple baseline, and add your own module to observe the shift.
• Change the loss, switch datasets, and see if something that worked for others still works for you.

Then you can decide whether this paper is worth reading up to 70%, 80%, or even 90%.

The reading order should be driven by need, not by guilt.

Reading, Just a Tool​

Reading papers is not a ritual, a competition, or a test of faith.

It’s just a tool, just a method, just a few stones you’ve accidentally picked up along the way.

Some stones you’ll use to build bridges, some you’ll leave behind, and some you won’t even notice as you step over them.

My notes are simply a record of the stones I’ve picked up.

Not every paper is fully understood, and not every paper is worth reading a second time.

But I try to write down the fragments that once stopped me, reminded me, or helped me rethink.

They’re not perfect, and they’re not great, but they happened.

If you’re still searching for your own reading rhythm, maybe some of these notes can help you avoid a few detours, or at least make you feel less alone.

Conclusion​

Reading papers is not about looking impressive.

It’s about having something in your mind that you can use when you need to make decisions.

As for the rest, leave it to gradient descent.

Because no matter what, it’s more reliable than fate.

☕ Fuel my writing with a coffee

Your support keeps my AI & full-stack guides coming.

All-in

AI / Full-Stack / Custom — All In

From idea to launch—efficient systems that are future-ready.

All-In Bundle

• Consulting + Dev + Deploy
• Maintenance & upgrades

🚀 Ready for your next project?

Need a tech partner or custom solution? Let's connect.