(@ junior researchers you should definitely read those)

Tomas is not on social media, so rare to meet a true superstar who stays away from spotlights, all the noise and just focused on doing solid work. Discussions with him were so refreshing and enlightening. A true visionary ... AI research needs more people like this!

⭐️(Poster) WildBench: Benchmarking LLMs with Challenging Tasks from Real Users in the Wild:
📜 arxiv.org/abs/2406.04770
⏰Th, Apr 24, 04:00 CEST

⭐️(Poster) Steering Masked Discrete Diffusion Models via Discrete Denoising Posterior Prediction:
📜 arxiv.org/abs/2410.08134
⏰Th, Apr 24, 04:00 CEST

⭐️(Oral) AI as Humanity's Salieri: Quantifying Linguistic Creativity of Language Models via Systematic Attribution of Machine Text against Web Text:
📜 arxiv.org/abs/2410.04265
⏰ Fr, Apr 25, 05:06 CEST

Thanks to Jekaterina Novikova and Malikeh Ehghaghi for inviting me. This is a wonderful initiative to encourage females to pursue a career in STEM.

Thanks to our speakers:
@nicolaschapados.bsky.social
@daniel-fried.bsky.social
@chrmanning.bsky.social
@sivareddyg.bsky.social
@yusu-agent.bsky.social
Tao Yu and Roberta Raileanu

Got ideas to share and want to learn about the latest progress?

Consider submitting your work! 🔗https://realm-workshop.github.io

Organizers:
@shikharmurty.bsky.social @ehsk0.bsky.social @xhluca.bsky.social @alex-lacoste.bsky.social @hanna-nlp.bsky.social @gneubig.bsky.social

“AI Agents” is a hot topic this year, but key issues remain such as reliability, generalization and robustness. How can we make progress towards these issues?

REALM’25 aims to bring researchers and practitioners interested in instruction-following grounded agents.

#ACL2025NLP

The papers presented were also so rich so please check them out. A huge thanks to everyone who attended and to my amazing co-organizers:

Watch video: neurips.cc/virtual/2024...

Francois Chollet talked about the lessons learnt from the ARC-AGI benchmark, test-time training, how to get LLM to adapt to novelty, etc

Jason Weston talked about self-training methods for equipping LLMs with System2 reasoning