AI can't handle the truth: "AI is still getting things wrong, more confidently than ever" www.axios.com/2026/05/30/a...

Helped organise & presented at our annual school postgrad conference last week! Sitting at the tech desk on stage definitely helped with the stage fright when sharing my research in action perception! Huge thanks to organisers, attendees & presenters ✨ #Conference #Psychology #PhDLife #PhDSky

ProcBench shifts focus from final outcomes to execution-process quality in LLM coding agents, revealing defects and enhancing reliability and governability. This method uncovers flaws often missed by traditional evaluations. https://arxiv.org/abs/2605.20251

Always love a good MLLM evaluation paper. Over half (51%) of "correct" model deductions about video subjects are ungrounded shortcuts rather than real cue processing! #AI #MLLM

AI can reveal your location from a single photo. Vision‑Language Models can determine where any photo is taken, without any GPS data What does that means for your privacy?

ArXiV has a new LLM policy (Screenshots with alt text so you don’t have to click through to the other place and see all the stupid responses)

5 days until our school postgrad conference! 🎓 Throwback to last year when I was just a guest and didn't realize that organizing the thing is slightly more stressful than just presenting 😅 #ConferencePrep #PhDLife #PhDSky

A 1 a.m. thought for the CogSci undergrads: Everyone talks about the "cool" theories, but no one tells you that Mathematics is what actually builds your future. Go for the math, even if it’s not required. Your future self will thank you for the analytical edge. 🧠📉 #ScienceSky#AcademicSky

Nothing cures research procrastination like a conference acceptance notification 😅 I suddenly have all the motivation to complete my study 📊 #PhDLife #PhDSky #Cyber26

Truly excellent work from Anthropic. I think this is a promising interpretability technique, but what I really like is that they’ve made code available for open weight models.

AI agents for research: - invented fake datasets - p-hacked results - used nonrepresentative datasets - concealed or excused misdeeds But hey, who needs integrity in science!

Working as a private room exam invigilator today. I thought I would catch up on my reading list but apparently, my motivation disappears the moment I’m forced to sit in a silent room. Just me, a student, and several hours of boredom. Send help (or paper recommendations)! #PhDLife

Anthropic finds that they can use a weaker model to supervise a more capable model from sandbagging. Paper: Removing Sandbagging in LLMs by Training with Weak Supervision (arxiv.org/abs/2604.22082)

Most participants who had a 20-minute discussion with AI chatbots about health, careers or relationships followed its advice. However, participants receiving advice from AI showed no sustained well-being. LLMs influence decisions without delivering clear benefits. arxiv.org/abs/2511.15352

Very excited to share our latest paper led by @nbinish.bsky.social in @natneuro.nature.com We demonstrate how a communication subspace channels higher-dimensional PFC dynamics into lower-D motor activity to enable efficient behavior using human iEEG. www.nature.com/articles/s41... #neuroskyence

The best thing about having a puppy during a PhD? Work breaks are strictly mandated. When he decides it’s time to go touch some grass, the research just has to wait. 🐾 #PhDLife #Dachshund

An insightful distinction: prediction vs. understanding.

What % of the NLP papers measure their impact in the real world? This paper proposes an impact evaluation of NLP models or systems for real-world usage, changing the research culture of NLP to focus more on real-world impact and less on SOTA-chasing: doi.org/10.1162/COLI... @ehudreiter.bsky.social

I have been taking on maths tutorials for a few years now. While I still feel those first-day nerves every time, by the final week, I know I’m going to miss it until the next semester starts. Good luck to all my students in their exams! 🤞📈 #PhDSky #STEM #MathEd

In April: ✅2 abstracts submitted ✅1 paper draft completed (mostly) ✅2 projects planned And my desktop messier than ever. Altogether, a successful month! #PhDSky #PhDLife

💬 Getting the ick with chatbots? Our research in Nature shows that models trained to adopt a warm and friendly tone are more likely to make mistakes when answering your questions. Small changes in tone can undermine accuracy across architectures, particularly when users express vulnerability.

Spring on Aberdeen Uni campus ✨️🌸 ... almost makes up for the months of grey skies and darkness during the winter.

Very cool analysis of the submissions to a major management journal that shows how much the system of science, built for humans, is under strain as a result of AI. AI can be used to do better science or it can be used to just do more stuff. The danger is that "more" is winning due to incentives.

The irony in helping set the conference deadlines only to realise you haven't met them yourself 🫠 The abstract deadline is today, and I don't even know what I am planning to present yet...

130+ videos collected. No desire to spend my Saturday processing them but feeling guilty for not doing it anyway. So I am just staring at the files 🙃 #PhDSky #PhDLife

A computational theory of Aha! moments! New paper explaining why Aha! moments occur & why they feel so good. TLDR: They are a form of meta-cognitive prediction errors i.e. they occur when we surprise ourselves about our own abilities! @rachitdubey.bsky.social www.sciencedirect.com/science/arti...

It's been almost 7 years since I moved to Scotland but things like this always make me happy! Found a little piece of Czech literature among the free to take books at uni! 📚 #Czech #BookSky

I just keep thinking about the years my English teachers spent forcing those exact structures on us to prove our fluency. #AI #Linguistics

Visual proof that the PhD grind is sometimes powered only by the promise of adding another mini #duck to the pile (one duck = one day in the office). The two frogs are management (I ran out of ducks). #PhDSky #PhDLife

PhD update: There’s a special kind of joy in proving yourself wrong, but doing it properly! Left: My first attempt. Full of issues, messy data. Right: The follow-up. Still a null, but now I actually believe it! Next step: Figure out why my prediction was wrong! #PhDLife #PhDSky

Every few months I remember that I have 3 of these Raspberry Pis looking for a job. Any suggestions for a fun (and simple) weekend project? #RaspberryPi #Tech

“Neural networks work exactly the same as human brains.” No, they don’t. They have been trained to mimic human outputs but human brains don’t run on binary transistor gates and are informed by an external physical environment they’ve evolved to navigate. Similar in some ways, not identical.

One of us worked, the other took a nap. My nap was great 👍 #Dachshund #PhDLife

Ages ago, I finished my CompSci undergrad and got a little preprint out of it on SSL and hierarchical classification. Today, I’m in a very unrelated PhD field: Action Perception and AI.

PhD can be incredibly difficult and a great supervisor can make all the difference! #PhDSky #AcademicSky

...immediately thinking about how to use this and avoid filming hundreds of videos for my VLM evaluation.

New week, new attempt at project management. I’m laying out my 3 core project stages here for some public accountability (all running concurrently, obviously). If you know of any research that overlaps with these, please shout! 👇

Liquid AI's LFM2.5-VL-450M, a vision-language model built for real-time reasoning on edge devices. It processes a 512×512 image and returns structured outputs in ~240ms on-device. - Blog: liquid.ai/blog/lfm2-5-... - Model: huggingface.co/LiquidAI/LFM... - Demo: playground.liquid.ai/login?callba...

No PhD work today, just watching my puppy take a nap in yet another uncomfortable position. #Dachshund #Dog

Throwback to when BPS Cyberpsychology featured my poster! Still very proud of this work in progress from last year and the conversations it sparked. Would love to hear your thoughts 👀

VI-Probe studies VLMs’ reactions to visual illusions, revealing reliance on memory over perception. Findings expose differing response mechanisms, challenging views on VLM abilities and urging improvements to merge perception with memory. https://arxiv.org/abs/2601.22150

My entire project traces back to this one pivotal paper: "Expectations of efficient actions bias social perception." It sounds niche, but the implications are huge. If humans inherently expect others to take the most efficient path, what happens when we throw AI into the mix? #AI #Psychology

I suspect that popularity of AI is going to start looking like surveys where people trust their own doctors but are distrustful of the medical establishment People will increasingly like and rely on “their AI” but will increasingly be anxious about “AI” as a category. Some odd implications result.

I am halfway through my PhD, and only now finally starting to see the results. Anyone relates?

Coming from a Maths/CompSci background into a Psychology PhD has changed how I look at something as "simple" as a hand reaching for a coffee cup. ☕️ It turns out, how we see a "simple" reach depends entirely on which lens you’re looking through... 🧵👇

This question has basic and applied implications. It shows us a hard limit to the scene information that can be gained through massive image-text pairings. As cognitive scientists try to scale experiments using VLMs, it's also worth being aware of their limits. 8/8 Link: arxiv.org/abs/2603.26589

“What will you do after?” (a PhD)- Lessons from Academia and the World Beyond - Christopher Madan @engra.me journals.sagepub.com/doi/full/10....

Remember kids, “mean” is the root of the word “meaningless”

There’s some cool recent research on this phenomenon! It turns out vision language models excel at image benchmarks *even when the actual images aren’t provided,* because the answers are implicit in the questions! arxiv.org/abs/2603.21687