Wednesday, November 2

 I woke up around 10:50am, even though the alarm was set to 10. I wish I could go to Europe and fix my drifting schedule a little. After Europe, my jet-lag becomes a good person's schedule: waking up at 6am, getting so many things done by 11. Anyway, in a week I'll get a free hour to correct my schedule a little.

Looking into Topology homework first after coffee to see if I better run to the office hours (one dumb question saves an hour). Then NLG paper. My model was overfitting; I did realize that there was no time for fixing it, but somehow I still tried these and those different learning parameters. I just ran it to see if I can get anything better. It's fine, it's just a midpoint checkup, but it'd be good to push it further. 

2-4pm NLG seminar: text summarization. Text summarization is what our professor studied as long time ago as in the 1980s, so she presented the topic with many anecdotes from her life. Text summarization is still in its early stages – factual constituency is a major issue (model "hallucinating" facts), and it's hard to generalize models to all topics the same way as large language models do now. Then we had three papers presented; all quite exciting. News summarization, entity-specific summarization (tell me what this long text say about a topic), and conversation summarization. Five years ago, I would be quite skeptical of this field, but now it's so much progress going on.

4-9pm writing my NLG paper. I got quite productive after drinking Red Bull at 5pm, wrote 5 pages. But I skipped the most important part – how do I adjust my algorithm, how do I fix it all – it remained unanswered. Some dudes next table were stinky – hmm no, I know we are all stressed these days, but showers give good ideas too and relax; it's worth taking them. A girl across my table was taking a nap on her laptop – that I approve, but I couldn't take a nap after Red Bull.

Then I had a dinner, chilled for an hour looking into the social media and other nonsense on my phone, then tried to write down the Topology homework. It seemed fine, but at some point my mind started melting and getting confused by elementary things. 2+2 were not 4 anymore, and it seemed that I missed something important in my life to realize that, and it was too late to figure it out. 

I went to bed by 3, but I couldn't fall asleep till about 4am. Maybe I shouldn't drink Red Bull in the late afternoon.