iiinova — where your digital self keeps evolving

aryanshri123.bsky.social

@aryanshri123.bsky.social

Reposted by @aryanshri123.bsky.social

Dang Nguyen @divingwithorcas.bsky.social · 26/09/2025

HR Simulator™: a game where you gaslight, deflect, and “let’s circle back” your way to victory. Every email a boss fight, every “per my last message” a critical hit… or maybe you just overplayed your hand 🫠 Can you earn Enlightened Bureaucrat status? (link below!)

245 5

Reposted by @aryanshri123.bsky.social

chenhaotan.bsky.social @chenhaotan.bsky.social · 09/07/2025

Prompting is our most successful tool for exploring LLMs, but the term evokes eye-rolls and grimaces from scientists. Why? Because prompting as scientific inquiry has become conflated with prompt engineering. This is holding us back. 🧵and new paper with @ari-holtzman.bsky.social .

23715 37

aryanshri123.bsky.social @aryanshri123.bsky.social · 03/07/2025

🤫Jailbreak prompts make aligned LMs produce harmful responses.🤔But is that info linearly decodable? ↗️We show many refused concepts are linearly represented, sometimes persist through instruction-tuning, and may also shape downstream behavior❗ arxiv.org/abs/2507.00239 🧵1/

161 6

aryanshri123.bsky.social @aryanshri123.bsky.social · 20/06/2025

We expose the "absence blindness" in the best LLMs, even when considering relatively short documents. Using LLMs as a judge or as graders may not be so reliable. Looking forward to see what comes out of this!

010 1

Reposted by @aryanshri123.bsky.social

Harvey Fu @harveyfu.bsky.social · 20/06/2025

LLMs excel at finding surprising “needles” in very long documents, but can they detect when information is conspicuously missing? 🫥AbsenceBench🫥 shows that even SoTA LLMs struggle on this task, suggesting that LLMs have trouble perceiving “negative spaces”. Paper: arxiv.org/abs/2506.11440 🧵[1/n]

27416 74