iiinova — where your digital self keeps evolving

Mariia Nebesnaia

@nemariia.bsky.social

44 followers 73 following 19 posts

PhD candidate & educator in Computer Engineering | AI & Optimization

Posts Replies Media

Reposted by Mariia Nebesnaia

Gaia Molinaro @gaiamolinaro.bsky.social · 08/10/2026

Our reward function compression work is now out in Nature Communications! 🎉 In it, we present and empirically test a computational framework that formalizes the trade-offs between goal flexibility and learning efficiency. @annecollins.bsky.social @natureportfolio.nature.com doi.org/10.1038/s414...

Conceptual model of goal-directed reinforcement learning with reward
function compression. In our proposed conceptual model, we break down goald-dependent learning into several parts. Combined with other learning components, an RL learning algorithm informs value estimates W on a given trial t, which are entered into a policy. The agent selects an action a through a joint policy, to which the environment responds by outputting outcome information o and a new state s. A working memory (WM) component organizes and maintains task-relevant information (here, broadly intended and dubbed the task information), used to support the reinforcement learning (RL) process. a Initially, value attribution to o is operated by WM,which maintains the current goal (g) and supports its matching to the experienced outcome to generate a reward r used by the RL agent. WM also holds a trace of recently experienced goals. b If goals share enough similar features,
a compressed reward function can be created, which replaces active goal maintenance and can, over time, be transferred to long-term memory storage. Eventually, the compressed reward function can be directly employed to assign value to experienced outcomes, sparing resources for other WM tasks.

08927

Mariia Nebesnaia @nemariia.bsky.social · 06/10/2026

For two months, I tried to surpass a specific value for some metric. Turned out it was mathematically impossible under my experimental setup. Have improved the algorithm greatly, though #highered #research

000

Mariia Nebesnaia @nemariia.bsky.social · 16/09/2026

At my main workplace, I became responsible for uploading videos to our department's YouTube channel. I want to make the channel interesting, so I have to convince them to let me do this too.

000

Mariia Nebesnaia @nemariia.bsky.social · 11/08/2026

The administration said that we shouldn't decrease grades even for obvious 100% AI phrases (which means a student just copied the AI response without checking what's inside). Well. What are we grading then and why #AI #highered

000

Mariia Nebesnaia @nemariia.bsky.social · 15/07/2026

My friend is doing research on the impact of AI use on English language learning. I wonder if people who learned it with AI will speak like AI #AI

010

Reposted by Mariia Nebesnaia

James MacGlashan @jmac-ai.bsky.social · 10/11/2024

Reinforcement learning in #AI is hard, so I’ve made a website to collect answers I’ve given to common RL questions. It's named Decisions & Dragons. It’s launching with 8 questions and answers, but I will add to it in the future. A 🧵 to give a preview with the link below.

3395

Mariia Nebesnaia @nemariia.bsky.social · 07/07/2026

I'm now officially a PhD candidate. I don't know why my department included 10 courses into this one exam, but I did it! Can finally focus on research. #highered #academiclife

000

Mariia Nebesnaia @nemariia.bsky.social · 15/04/2026

In programming languages, we have parameter passing by value, by reference, by result, etc. And, of course, by ✨️magic✨️ #CSed #highered

A slide from a lecture on programming languages: parameter passing by magic

010

Mariia Nebesnaia @nemariia.bsky.social · 19/02/2026

I'm not against students using AI - I want them to use it well. But many struggle to: prompt effectively, evaluate outputs critically, debug AI-generated code, or understand what they're submitting. AI is here to stay, and we have to teach AI literacy. #CSed #AI

010

Mariia Nebesnaia @nemariia.bsky.social · 12/11/2025

Been thinking about sustainable creator monetization models that don't rely on intrusive ads. Direct support platforms require audience brought from outside. What would an optimization model look like that balances discovery, creator support, and user experience? Is there a sustainable solution?

100

Mariia Nebesnaia @nemariia.bsky.social · 31/10/2025

Students have great mobile and social media skills but limited desktop/CLI experience. I think it's time to make not only web apps mobile first but also our educational programs. #CSed #academiclife

000

Reposted by Mariia Nebesnaia

LMU Open Science Center @lmu-osc.bsky.social · 23/10/2025

🌍 During our Open Science Summer School, Prof. Dr. Felix Schönbrodt's (@nicebread.bsky.social) lecture “Open Access, Preprints, Postprints - or ‘How to do Open Access For Free’,”🔓 explored ways researchers can control how their work is shared. Full lecture: osf.io/5rsxb; Slides: osf.io/c2mrb

A close-up of a hand reaching towards books on a shelf. The sleeve of a beige pullover is visible. The open access “unlock” symbol is overlaid transparently on the image.

163

Mariia Nebesnaia @nemariia.bsky.social · 24/10/2025

Teaching a security lab on computers without root... Got to be creative. Ended up using docker containers #infosec #academiclife

000

Mariia Nebesnaia @nemariia.bsky.social · 13/10/2025

I'm in love with direct #p2p and local network discovery apps. Currently building one. Some of my colleagues didn't know that's even possible! I think it's so important to talk about these technologies

010