🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

Excited to be in Singapore for ICLR 2025! 🇸🇬 📷We will present BALROG at the poster session on Saturday, 3:00-5:30 PM, Hall 3, #252 Sneak peak at the poster, including the updated leaderboard with some new models, more on them soon 👀 Bartłomiej Cupiał, Ulyana Piterbarg, Tim Rocktäschel

Presenting this today 3-530 at poster #208, come say hi 🙋‍♀️

Very excited for this keynote by Tim Rocktäschel! Awesome to see open-endedness go from a niche (😉) area to a keynote at #ICLR ! 🌱🌿🌳🌲🍀🌍✨ 📈 🧬🧪 cc Joel Lehman Kenneth Stanley

Awesome to see a keynote on open-endedness at #ICLR - way to go Tim Rocktäschel ! You have the right message at the right time and I appreciate the callout in the abstract. I wish I was there to see this. Open-endedness is the next frontier for AI as the benchmark race loses its allure.

🚨 New top scorer on the BALROG LLM leaderboard! DeepSeek R1, evaluated via NVIDIA’s NIM API, takes the lead. Its strong reasoning capabilities help it make solid progress on BALROG’s toughest tasks — the best performance we’ve seen so far! 🧠⚔️

Evaluating DeepSeek R1 on BALROG wouldn’t have been possible on our academic budget. Huge thanks to NVIDIA NIM, Karin Sevegnani, and Georg Zitzlsberger for making it happen! We’ve shared more details in this blog post — check it out: developer.nvidia.com/blog/benchmark…

👑 BALROG has a new King! 🥊 Gemini-2.5-Pro enters the ring and dominates its competitors, outperforming the second-best model by ~6%.

Not only is Gemini-2.5-Pro currently the top performer, it’s also by far the most cost-effective model in its performance category!

Join UCL DARK's Tim Rocktäschel for a can't-miss keynote diving deep into how Open-Endedness, World Models, and Automated Innovation are going to shape the next wave of AI. 🔥 #ICLR2025 🇸🇬 🕑 Starting at 2PM (in 30 minutes) 📍 Hall 1 Apex

Tim Rocktäschel’s keynote talk at #ICLR2025 about Open-Endedness and AI. “Almost no prerequisite to any major invention was invented with that invention in mind.” “Basically almost everybody in my lab at UCL and at DeepMind have read this book: Why Greatness Cannot Be Planned.”

Tim Rocktäschel - congratulations on such an exciting ICLR 2026 keynote! A case for world models as step to open ended agent learning, including robotics

Such an inspiring and thought-provoking keynote talk by ⁦⁦⁦Tim Rocktäschel⁩ at #ICLR2025 highlighting a promising path towards automating innovation: combining foundation models with open-endedness methods 🔥

Incredible keynote by Tim Rocktäschel! Some might fear that open-endedness is a disease for AI safety; I say it might be the remedy. Our 🌈Rainbow Teaming work proved these methods are great at finding & fixing LLM vulnerabilities. Hope the safety folks use more of these powerful methods

Excited to share our progress in scaling foundation world models with a packed room at ICLR 2026

What an amazing and thought-provoking talk by Roberta Raileanu. Despite all the recent progress, we’re still far from AI agents making true scientific breakthroughs — and there’s so much important work ahead.

Roberta Raileanu made a strong case today in her talk. No need to worry about AI saturating research — there are still plenty of deep, open problems. This slide captures it well and she shared many interesting directions worth exploring!

Indeed, plenty of open research problems to work on, despite the hype of AI automating all AI research.

This got accepted to #ICML2025 as a *spotlight paper* (top 2.6%!) 🚀 --- work that Yi Xu did as an Msc student! Surely this will mark the start of an exceptional academic journey

Our UCL DARK MSc student Yi Xu managed to get his work accepted as a spotlight paper at ICML Conference 2025 (top 2.6% submissions) 🚀 What an amazing success testament to the outstanding supervision by Robert Kirk and Laura Ruis.

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