General Update
Intro
My last blog was after the entertaining wimbledon final last year, and honestly it doesn’t feel like some 9 months have passed. If anything I think it’s more of an indicator of me being preoccupied with other stuff (mostly just research), although a secondary reason was that I really didn’t travel that much. In fact, the only reason I am writing this is as temperature keeps rising, in true TSRB fashion, the lab becomes basically an oven without air conditioning over the weekend. During the past several hours, I’ve already tried to get any work done but ultimately in vain due to the insane heat. Alas, I decided to use my time somewhat productively, in doing something that requires no concentration.
There were several instances where I felt like writing down a blog, one such instance was the tragic passing of GM Daniel Naroditsky. I do not wish to discuss the event anymore but in remembrance of someone quite important to my chess journey, I’d like to open the blog with an offbeat line in the Najdorf Sicilian (possibly my favorite), which I came across in one of Danya’s speedrun videos.
Sicilian, Open, Najdorf, English Attack
This line is one of the sidelines of English Attack of the Najdorf Sicilian, which arises after
1. e4 c5
2. Nf3 d6
3. d4 cxd4
4. Nxd4 Nf6
5. Nc3 a6
6. Be3 e5
7. Nb3 Be6
8. f3 Be7
9. Qd2 O-O
10. O-O-O a5?!
One can also potentially arrive at this position from a different move order after 6.f3. And in my opinion, around 2000 blitz/bullet, this is the second most common position from Najdorf only after 6.Bg5. From this position, white is looking for g4 g5 and rapid expansion on king side, since black has already committed to castle, as well as trading on d5 and putting a pawn there gaining space. The general game plan for black is to develop the b knight, looking for the d-file pawn break as (one of the weakness of the move 6.e5 is to create a weakness on d5 square, compared to the 6.e6 line in many variations), and put the queen or rook on c file, creating pressure against the king. The most common moves for black are now 10.Nbd7 or 10.b5, very logical.
However, the move 10.a5?! is an offbeat sideline for black, with which I’ve had great success with. This is not objectively a good move, however it is also difficult to refute at 2000 level, without previously known such a line. It looks silly at first since black has already committed a6, seemingly wasting a tempo, however, the idea becomes apparent if opponent does not find the optimal contiunation. The most common response from white at my level (2000 blitz/bullet chess.com) is 11. g4, hoping to continue with the English attack. This move, however, already puts white under disadvantage, because after
1. O-O-O a5?!
2. g4 a4!
3. Na1 a3
4. b3
White’s knight is forced to retreat to the corner while having a huge weakness at b2 square, also weakening the c3 knight and a common winning construction involves puting bishop or queen on the long diagonal. For anything other then 11.a4 or 11.Bb5 from white, a4,a3 is pretty much always good for black. In the above position, there is often an idea with Ra5, triple defending d5 square, making the pawn break possible and putting a rook on d5 square, forcing white’s queen to sidestep.
The most challenging move for black is 11.a4, stopping any further pawn push. However, this weakens the b4 square, and black puts a knight on b4, again triple defending d5 square and looking for pawn push. Though this line is often empirically very dangerous for white at this level, since the a-pawn often becomes undefended after we challenge the c3 knight, creating weakness around the king. I think this is a very good practical line and I recommend to try it out.
Research
ITW 2025
So in early September last year, ITW happened but I was not able to go due to Visa issues. The conference was also in Australia so Dr. Bloch presented on my behalf. There isn’t much to say check out my work. We’ve got some expected pushbacks from communication theory people and I think looking forward I am going to largely refrain from doing exponent stuff because I really dislike it as a metric. But I am not going to spend too much time on it in this blog. I am looking probably to extend the conference with a security angle but currently it is not on my pipeline at all.
ISIT 2026
I also submitted to ISIT this January and got accepted last week. And surprisingly got very positive reviews considering I had cut my converse proof from the conference version as I could not remedy a last minute mistake. To be fair I do think it was a good paper but not able to finishing the converse as well a bound on Chernoff exponent between mixtures really frustrated me at some point so I am also pretty glad it went well. I will probably be extending this to a journal once I fixed my converse, although it is also not currently on my pipeline.
The paper was really about a security angle on the recently popular integrated sensing and communication area. We had argued that a similary dual-purpose adversary will require transmitter to handle communication security and sensing privacy very differently and very few papers have realized that one usually cannot guarantee both. The paper is currently on arxiv so please check it out.
Covert LLM Embedding
The project that I spent most of my (actually all my) time for the past several months is not related to communication at all. Although the techniques used are largely similar to what one would use for a communication coding problem. Specfically, I am working on a information-theoretic formulation (as well as designing a practical algorithm) for multi-bit covert watermarking in LLM. The capacity formulations are largely done but I am currently implementing an algorithm that I think will beat basically every other state-of-the-art multi-bit watermarking in terms of embedding rate. The goal of the project is really to provide a mathematical formulation for an otherwise very algorithmic problem. The CS and crypto community handles such problems very differently, and a lack of rigid modeling and upper limit result really convinces us to explore from an information-theoretic perspective. I think I am in good shape to submit to ITW2026 again, but one can never be sure for a paper that requires implementation results.
Travels and future plans
I did have a short travel towards the end of last month but there wasn’t anything worth nothing. One thing worth noting was that my chess and tennis skill improved quite a bit and I think in general I am in better physical shape, which one do not expect during a PhD.
This is probably my last update before fall semester as, unless something happens, I am traveling to Boston this summer as a visiting scholar to collaborate with Dr. Flavio Calmon at Harvard, in, once again, watermarking in LLM. His works have always impressed me as he is one of the few people in information-theory communication that applies it to very practical researches in the broad areas of security and privacy so I am definitely looking forward to it. Also I have to say I’ve become quite sick of TSRB I do need somewhat of an environement shift at this point.
There are also several things in my pipeline that I’ve been postponing, one of which is to improve my research workflow by designing several custom programs, I feel like I really need to renew my Quarto markdown template for presentation as well as a better latex render that allows one to edit parts of an equation in real time. But I have not been in urgent need for those recently, though I do wish to finish these personal projects this year.
Some final thoughts
One thing that I will update was my Yorushika blog as there is now a new Yorushika Album, and although I haven’t mentioned, there is also a new Album from Zutomayo, who has recently become a close second in terms of music likings.
Coincidentally, chess candidate 2026 is currently ongoing and as of me writing this, Sindarov is on 5.5/6. As a Fabi supporter I am alsolutely pissed after what happened for like last three candidate. So expect a blog if Fabi somehow wins, otherwise I pretend this candidate never happened.
Enjoy Reading This Article?
Here are some more articles you might like to read next: