Scientia AI●

a bilingual, research-first reading log for AI

auth

Tools

Cite your source correctly and post your Weekly Reading Discussion response. No sign-in needed.

How to use this page — 4 steps

  1. Get your paper's BibTeX or DOI (from Google Scholar, arXiv, or the paper's page). Paste it below and click Convert.
  2. Click "↓ Insert IEEE citation" to copy it into the post template.
  3. Fill in the blanks. Look at the example post below if you're not sure what to write — simple sentences are fine.
  4. Click Copy template, then paste it into your GitHub Discussion post.

Citation converter

Paste a BibTeX entry (text starting with "@") or just a DOI (an ID number like 10.1000/xyz123). Choose a citation style below. Use IEEE for the Weekly Reading Discussions.

Weekly Reading Discussion

One discussion thread per week. Post your own reflection first, then leave 2–3 comments on classmates' threads.

Go to this week's Discussion →

How it's graded

A summary is not enough — any AI tool can write a summary. Every post needs all 3 parts:

  1. The paper's IEEE citation.
  2. A short summary (1–2 sentences).
  3. One specific thing, in your own words: something you didn't know, a weak point you noticed, or an idea for future research.

About AI

Key words

Citation
A short reference to a paper — who wrote it, where, and when — so a reader can find it.
DOI
"Digital Object Identifier" — a permanent ID number for a paper, like 10.1000/xyz123.
BibTeX
A citation format many databases (Google Scholar, Zotero) let you export. Starts with "@".
Reflection
Your own thoughts about the paper — not a summary of what it says.
Disclose
To say clearly and openly. Example: "I used AI to check my grammar" is a disclosure.
et al.
Latin for "and others" — used when a paper has too many authors to list them all.

Your first post — template

Fill in the blanks below (see the example underneath if you're stuck), then copy the whole thing into your Discussion post.

Example of a finished post

This is a made-up example, just to show you the pattern — don't copy the words, write your own.

K. He, X. Zhang, S. Ren, and J. Sun, "Deep Residual Learning for Image Recognition," _Proc. IEEE Conf. Comput. Vis. Pattern Recognit._, pp. 770–778, 2016.

**AI-assisted:** grammar & wording check only

**Summary:** This paper introduces "residual learning," a way to train much deeper neural networks by letting each layer learn a small change instead of a whole new representation.

**Something I didn't know before reading this:** I didn't know that simply stacking more layers onto a normal network can make it perform *worse*, not better — I always assumed deeper meant more accurate. The "shortcut connections" that skip layers were the part that surprised me most.

**A question or note for future research:** The paper doesn't test networks deeper than 152 layers. I'd be curious whether the same trick still helps at 1,000+ layers, or whether some other problem takes over at that point.

Replying to classmates — two samples

Leave 2–3 of these on classmates' threads each week. Reply directly under their comment, so it's clear who you're talking to.

Sample 1 — build on their point

Sample 2 — respectful pushback

AI-phrasing self-check

Paste your draft below and check it before you post. This is a simple checklist tool, not an AI detector — it does not prove anything. It just checks for generic phrasing and the personal observation the assignment asks for. Note for non-native speakers: formal English you learned for essays or TOEIC is completely fine — this tool is not judging your English level.