Davide Paglieri

LLMs acing math olympiads? Cute. But BALROG is where agents fight dragons (and actual Balrogs)🐉😈 And today, Grok-4 (@grok) takes the gold 🥇 Welcome to the podium, champion!

Grok 4 results on @NetHack_LE just dropped!

1 43.6 Grok-4-Wiz-AI-Cha died in The Dungeons of Doom on level 1. Killed by a housecat.

Finally a high score we can be proud of.

No worries 😉

The world is moving towards agents Static benchmarks don't measure what agents do best (multi-turn reasoning) Thus, interactive benchmarks: * Terminal Bench (@alexgshaw, @Mike_A_Merrill) * Text Arena (@LeonGuertler) * BALROG (@PaglieriDavide, @_rockt) * ARC-AGI-3 (@arcprize)

💯 Who knew that the International Math Olympiad (IMO) is much easier than @NetHack_LE for AI.

Official results are in - Gemini achieved gold-medal level in the International Mathematical Olympiad! 🏆 An advanced version was able to solve 5 out of 6 problems. Incredible progress - huge congrats to @lmthang and the team! deepmind.google/discover/blog/…

I think ARC is a great eval, but at this point we should just use nethack

[6/n] Check out the tech report: arxiv.org/abs/2507.08892

Introducing Concordia 2.0, an update to our library for building multi-actor LLM simulations!! 🚀 We view multi-actor generative AI as a game engine. The new version is built on a flexible Entity-Component architecture, inspired by modern game development.

In May I missed a single email from openreview saying I'd be auto-enlisted as a reviewer. Then a few ACs missed my immediate and repeated messages on openreview saying that I won't be able to review since I'll be taking the second half of my paternity leave. Now all of my…

The race for LLM "cognitive core" - a few billion param model that maximally sacrifices encyclopedic knowledge for capability. It lives always-on and by default on every computer as the kernel of LLM personal computing. Its features are slowly crystalizing: - Natively multimodal…

Some personal news: I joined Google DeepMind in @_rockt's uber talented Open-Endedness team. I couldn’t be more excited for what we’re cooking. AI is the least open-ended it will ever be. Meta, it’s been a blast, an honor, and a privilege. I’m very grateful for the freedom and…

**When AIs Start Rewriting Themselves** Darwin G​​ödel Machine: Open-Ended Evolution of Self-Improving Agents The Darwin G​​ödel Machine can: 1. Read and modify its own code 2. Evaluate if the change improves performance 3. Open-endedly explore the solution space 🧵👇

Excited to announce that this fall I'll be joining @jacobandreas's amazing lab at MIT for a postdoc to work on interp. for reasoning (with @ev_fedorenko 🤯 among others). Cannot wait to think more about this direction in such a dream academic context!

Gemini 2.5 Pro completes Pokémon Blue 🤯🔥 But how does it fare in much harder, more unforgiving games? On NetHack, it barely scratches the surface—just 1.7% progression, as tested in BALROG, our new benchmark for agentic LLMs 🗡️ Check it out: balrogai.com