Ilija Bogunovic

University of Basel@University of Basel·瑞士·OpenReview

研究方向

Bandit optimization · Reinforcement learning · Sequential decision making · Reinforcement Learning from Human Feedback · Reinforcement learning for LLMs · Post-training of diffusion LLMs

6.7

wd1: Weighted Policy Optimization for Reasoning in Diffusion Language Models

ICLR 2026Poster

通讯

5.0

This Is Your Doge, If It Please You: Exploring Deception and Robustness in Mixture of LLMs

ICLR 2026Rejected

三作

4.0

Synthetic Data is Sufficient for Zero-Shot Visual Generalization from Offline Data

ICLR 2025Rejected

三作

5.5

Right Now, Wrong Then: Non-Stationary Direct Preference Optimization under Preference Drift

ICML 2025Poster

通讯

4.7

Right Now, Wrong Then: Non-Stationary Direct Preference Optimization under Preference Drift

合作者 (20)

Ilija Bogunovic

wd1: Weighted Policy Optimization for Reasoning in Diffusion Language Models

This Is Your Doge, If It Please You: Exploring Deception and Robustness in Mixture of LLMs

RSPO: Regularized Self-Play Alignment of Large Language Models

Robust Multi-Objective Controlled Decoding of Large Language Models

Overton Pluralistic Reinforcement Learning for Large Language Models

Imagined Autocurricula

Sample Efficient Preference Alignment in LLMs via Active Exploration

Synthetic Data is Sufficient for Zero-Shot Visual Generalization from Offline Data

Right Now, Wrong Then: Non-Stationary Direct Preference Optimization under Preference Drift

Right Now, Wrong Then: Non-Stationary Direct Preference Optimization under Preference Drift