๐๐ Excited to announce that our paper โWhat if Othello-Playing Language Models Could See?โ is accepted at #EMNLP2025 Findings!
โญ๏ธ We explore how multimodal training boosts sample efficiency & board-state understanding using Othello as a testbed.
๐ Preprint coming soon!
๐๐ Excited to announce that our paper โWhat if Othello-Playing Language Models Could See?โ is accepted at #EMNLP2025 Findings!
โญ๏ธ We explore how multimodal training boosts sample efficiency & board-state understanding using Othello as a testbed.
๐ Preprint coming soon!
Excited to present at the #ICML2025 World Models Workshop!
๐ July 18, 15:45โ17:00
๐ง What if Othello-Playing Language Models Could See?
We show that visual grounding improves prediction & internal structure.
โ๏ธ๐ท๐ง
#ICML2025#AI#LLMs
Wrapping up at #EMNLP2024 in Miami!๐๏ธ Many thanks for the attendees and their feedback in my poster session. Had a great time joining oral/poster sessions and meeting amazing people! So much to learn and share! Ending with the #UvA group photo to remember this amazing experience!
๐ง Can LLMs actually reason? We have surprising observations on simple sequential reasoning tasks! Our paper in #EMNLP2024 Findings introduces the SIFo Benchmarkโa tool testing LLMs' ability to follow & reason over multiple instructions. ๐ Join our poster on Nov 12, 16:00-17:30!
๐ Our paper reveals that, while LLMs handle individual tasks, they struggle with reasoning over sequential instructions, highlighting a crucial gap in their robustness. ๐จ
๐ https://t.co/Avfm73fr6o
Tomorrow (Friday, Dec 8), I will present my poster atย #EMNLP2023 during the Language Grounding to Vision, Robotics and Beyond session in East Foyer from 11:00 - 12:30. ย Feel free to swing by!
๐ Exciting News! Our paper "The BLA Benchmark: Investigating Basic Language Abilities of Pre-Trained Multimodal Models" has been accepted at #EMNLP2023 Main Conference. Many thanks to my co-authors @sandropezzelle and @raquel_dmg.
#emnlp#NLP
Good taste:
1. Smart enough: Good & innovative ideas instead of what everyone else is doing?
2. Good research: Not something A+B, but have them with good reasons.
3. Propose something with a reason but not simply trying something popular.