Piotr Miłoś
Scholar

Piotr Miłoś

Google Scholar ID: Se68XecAAAAJ
Google, University of Warsaw, Polish Academy of Sciences
reasoning and controlreinforcement learningLLMsprobability theory
Homepage↗Google Scholar↗
Citations & Impact
All-time
Citations
2,738
 
H-index
19
 
i10-index
30
 
Publications
20
 
Co-authors
0
 
Contact
CVOpen ↗TwitterOpen ↗GitHubOpen ↗LinkedInOpen ↗
Publications
5 items
Up and Down the Abstraction Ladder: Code-Based Skills for Language Agents
2026
Cited
0
What Drives Interactive Improvement from Feedback?
2026
Cited
0
When Does Non-Uniform Replay Matter in Reinforcement Learning?
2026
Cited
0
Amortized Causal Discovery with Prior-Fitted Networks
2025
Cited
0
Joint MoE Scaling Laws: Mixture of Experts Can Be Memory Efficient
2025
Cited
0
Resume
Academic Achievements
  • Selected Publications:
  • - Contrastive Representations for Temporal Reasoning (NeurIPS 2025)
  • - Joint MoE Scaling Laws: Mixture of Experts Can Be Memory Efficient (ICML 2025)
  • - Since Faithfulness Fails: The Performance Limits of Neural Causal Discovery (ICML 2025)
Research Experience
  • Staff SWE at Google, working in the Cloud AI team.
Background
  • Machine Learning Researcher; Associate Professor at the Polish Academy of Sciences and University of Warsaw; Member of the ELLIS Society; Staff SWE at Google working in the Cloud AI team; Interested in methods that can deliver robust decision-making capabilities in complex scenarios.
Miscellany
  • Supported by the Polish Children's Fund; Quick links: Twitter, LinkedIn, CV