Lucia Cipolina Kun
@luciakun.bsky.social
Currently LLM agents at META.
AI researcher in Reinforcement Learning, LLMs and Cultural Heritage.
https://lcipolina.github.io/
AI researcher in Reinforcement Learning, LLMs and Cultural Heritage.
https://lcipolina.github.io/
Thank you Sacha!
August 16, 2025 at 9:40 PM
Thank you Sacha!
Work done during my postdoc at @fz-juelich.de with @jeniajit.bsky.social and Marianna Nezhurina. Thanks to the Open Spiel developers @sharky6000.bsky.social @shagoon.bsky.social and John Shultz.
August 16, 2025 at 10:06 AM
Work done during my postdoc at @fz-juelich.de with @jeniajit.bsky.social and Marianna Nezhurina. Thanks to the Open Spiel developers @sharky6000.bsky.social @shagoon.bsky.social and John Shultz.
🚀 Try it now:
📖 Blog: laion.ai/blog/reasoni...
📋 Paper: arxiv.org/abs/2508.03368
💻 Code: github.com/SLAMPAI/game...
🔬 Colab demo: github.com/SLAMPAI/game...
📚 Docs: game-reasoning-arena.readthedocs.io
🤗 Leaderboard: huggingface.co/spaces/lcipo...
📖 Blog: laion.ai/blog/reasoni...
📋 Paper: arxiv.org/abs/2508.03368
💻 Code: github.com/SLAMPAI/game...
🔬 Colab demo: github.com/SLAMPAI/game...
📚 Docs: game-reasoning-arena.readthedocs.io
🤗 Leaderboard: huggingface.co/spaces/lcipo...
Game Reasoning Arena: Inside the Mind of AI: How LLMs Think, Strategize, and Compete in Real-Time | LAION
<h3><a id="access" class="anchor" href="#access" aria-hidden="true"><svg aria-hidden="true" class="octicon octicon-link" height="16" version="1.1" viewbox="0...
laion.ai
August 16, 2025 at 10:04 AM
🚀 Try it now:
📖 Blog: laion.ai/blog/reasoni...
📋 Paper: arxiv.org/abs/2508.03368
💻 Code: github.com/SLAMPAI/game...
🔬 Colab demo: github.com/SLAMPAI/game...
📚 Docs: game-reasoning-arena.readthedocs.io
🤗 Leaderboard: huggingface.co/spaces/lcipo...
📖 Blog: laion.ai/blog/reasoni...
📋 Paper: arxiv.org/abs/2508.03368
💻 Code: github.com/SLAMPAI/game...
🔬 Colab demo: github.com/SLAMPAI/game...
📚 Docs: game-reasoning-arena.readthedocs.io
🤗 Leaderboard: huggingface.co/spaces/lcipo...
🛠️ Technical highlights:
• Built on OpenSpiel framework
• 9+ strategic games (Tic-Tac-Toe → Chess)
• Support for LiteLLM, Hugging Face and vLLM
• Ray parallelization for large-scale experiments
• Automatic reasoning categorization
• TensorBoard integration
Scale from single games to tournaments 🏆
• Built on OpenSpiel framework
• 9+ strategic games (Tic-Tac-Toe → Chess)
• Support for LiteLLM, Hugging Face and vLLM
• Ray parallelization for large-scale experiments
• Automatic reasoning categorization
• TensorBoard integration
Scale from single games to tournaments 🏆
Game Reasoning Arena: Inside the Mind of AI: How LLMs Think, Strategize, and Compete in Real-Time | LAION
<h3><a id="access" class="anchor" href="#access" aria-hidden="true"><svg aria-hidden="true" class="octicon octicon-link" height="16" version="1.1" viewbox="0...
laion.ai
August 16, 2025 at 10:04 AM
🛠️ Technical highlights:
• Built on OpenSpiel framework
• 9+ strategic games (Tic-Tac-Toe → Chess)
• Support for LiteLLM, Hugging Face and vLLM
• Ray parallelization for large-scale experiments
• Automatic reasoning categorization
• TensorBoard integration
Scale from single games to tournaments 🏆
• Built on OpenSpiel framework
• 9+ strategic games (Tic-Tac-Toe → Chess)
• Support for LiteLLM, Hugging Face and vLLM
• Ray parallelization for large-scale experiments
• Automatic reasoning categorization
• TensorBoard integration
Scale from single games to tournaments 🏆
🎯 Strategic specialization: Some LLMs excel at tactics but struggle with long-term planning
🎲 Reasoning diversity: Models show distinct decision-making patterns for different game
🔄 Game adaptation: Bigger models adapt their strategy as the game progresses.
🎲 Reasoning diversity: Models show distinct decision-making patterns for different game
🔄 Game adaptation: Bigger models adapt their strategy as the game progresses.
August 16, 2025 at 10:02 AM
🎯 Strategic specialization: Some LLMs excel at tactics but struggle with long-term planning
🎲 Reasoning diversity: Models show distinct decision-making patterns for different game
🔄 Game adaptation: Bigger models adapt their strategy as the game progresses.
🎲 Reasoning diversity: Models show distinct decision-making patterns for different game
🔄 Game adaptation: Bigger models adapt their strategy as the game progresses.
🔍 LLMs exhibit "reasoning entropy" and game-specific intelligence - adaptive vs. rigid strategic patterns revealed
🧠 What makes this unique? We capture the reasoning behind every move and analyse the strategies as the game evolves
Not just "what did the AI choose" but "why did it choose that?"
🧠 What makes this unique? We capture the reasoning behind every move and analyse the strategies as the game evolves
Not just "what did the AI choose" but "why did it choose that?"
August 16, 2025 at 10:01 AM
🔍 LLMs exhibit "reasoning entropy" and game-specific intelligence - adaptive vs. rigid strategic patterns revealed
🧠 What makes this unique? We capture the reasoning behind every move and analyse the strategies as the game evolves
Not just "what did the AI choose" but "why did it choose that?"
🧠 What makes this unique? We capture the reasoning behind every move and analyse the strategies as the game evolves
Not just "what did the AI choose" but "why did it choose that?"
Thank you Sacha! yes I will!
August 16, 2025 at 9:54 AM
Thank you Sacha! yes I will!
Congratulations Luke!
February 18, 2025 at 10:57 PM
Congratulations Luke!
Felicitaciones Pablo!
December 30, 2024 at 8:07 PM
Felicitaciones Pablo!
Thank you Sacha!
December 30, 2024 at 2:58 PM
Thank you Sacha!
Jenia Jitsev and Mariana Nezurinha are there. I can make the connections if you want :)
December 13, 2024 at 9:16 PM
Jenia Jitsev and Mariana Nezurinha are there. I can make the connections if you want :)
My lab will be presenting this paper on LLM reasoning
openreview.net/forum?id=Mkl...
On this workshop
scienceofdlworkshop.github.io
openreview.net/forum?id=Mkl...
On this workshop
scienceofdlworkshop.github.io
Alice in Wonderland: Simple Tasks Reveal Severe Generalization and...
Large Language Models (LLMs) are often described as being instances of foundation models - that is, models that possess strong generalization and therefore transfer robustly across various tasks...
openreview.net
December 12, 2024 at 2:13 PM
My lab will be presenting this paper on LLM reasoning
openreview.net/forum?id=Mkl...
On this workshop
scienceofdlworkshop.github.io
openreview.net/forum?id=Mkl...
On this workshop
scienceofdlworkshop.github.io