misovalko/my-research-papers

Michal Valko - Research Papers Collection

This dataset contains references to research papers by Michal Valko and collaborators.

Papers

• RL-finetuning LLMs from on- and off-policy data with a single algorithm
• Preference optimization with multi-sample comparisons
• Optimal design for reward modeling in RLHF
• A new bound on the cumulant generating function of Dirichlet processes
• Understanding the performance gap between online and offline alignment algorithms
• Sharp deviations bounds for Dirichlet weighted sums with application to analysis of Bayesian algorithms
• KL-entropy-regularized RL with a generative model is minimax optimal
• Accelerating Nash learning from human feedback via Mirror Prox
• The Llama 3 herd of models
• Metacognitive capabilities of LLMs: An exploration in mathematical problem solving
• Local and adaptive mirror descents in extensive-form games
• Nash learning from human feedback
• Human alignment of large language models through online preference optimisation
• Generalized preference optimization: A unified approach to offline alignment
• Decoding-time realignment of language models
• A general theoretical paradigm to understand learning from human preferences
• Unlocking the power of representations in long-term novelty-based exploration
• Demonstration-regularized RL
• Model-free posterior sampling via learning rate randomization
• Curiosity in hindsight: Intrinsic exploration in stochastic environments
• VA-learning as a more efficient alternative to Q-learning
• Fast rates for maximum entropy exploration
• Adapting to game trees in zero-sum imperfect information games
• Understanding self-predictive learning for reinforcement learning
• DoMo-AC: Doubly multi-step off-policy actor-critic algorithm
• Regularization and variance-weighted regression achieves minimax optimality in linear MDPs: Theory and practice
• Quantile credit assignment
• Half-Hop: A graph upsampling approach for slowing down message passing
• BYOL-Explore: Exploration by bootstrapped prediction
• Optimistic posterior sampling for reinforcement learning with few samples and tight guarantees
• From Dirichlet to Rubin: Optimistic exploration in RL without bonuses
• Retrieval-augmented reinforcement learning
• Scaling Gaussian process optimization by evaluating a few unique candidates multiple times
• Large-scale representation learning on graphs via bootstrapping
• Adaptive multi-goal exploration
• Marginalized operators for off-policy reinforcement learning
• Drop, Swap, and Generate: A self-supervised approach for generating neural activity
• Stochastic shortest path: minimax, parameter-free and towards horizon-free regret
• A provably efficient sample collection strategy for reinforcement learning
• Model-free learning for two-player zero-sum partially observable Markov games with perfect recall
• Unifying gradient estimators for meta-reinforcement learning via off-policy evaluation
• Broaden your views for self-supervised video learning
• UCB Momentum Q-learning: Correcting the bias without forgetting
• Fast active learning for pure exploration in reinforcement learning
• Revisiting Peng's Q(λ) for modern reinforcement learning
• Taylor expansion of discount factors
• Online A-optimal design and active linear regression
• Kernel-based reinforcement Learning: A finite-time analysis
• Game plan: What AI can do for football, and what football can do for AI,
• A kernel-based approach to non-stationary reinforcement learning in metric spaces
• Episodic reinforcement learning in finite MDPs: Minimax lower bounds revisited
• Adaptive reward-free exploration
• Fast sampling from β-ensembles
• Mine Your Own vieW: Self-supervised learning through across-sample prediction,
• Bootstrap Your Own Latent: A new approach to self-supervised learning
• BYOL works even without batch statistics
• Improved sample complexity for incremental autonomous exploration in MDPs
• Sampling from a k-DPP without looking at all items
• Statistical efficiency of Thompson sampling for combinatorial semi-bandits
• Planning in Markov decision processes with gap-dependent sample complexity
• Monte-Carlo tree search as regularized policy optimization
• Taylor expansion policy optimization
• Gamification of pure exploration for linear bandits
• No-regret exploration in goal-oriented reinforcement learning
• Improved sleeping bandits with stochastic action sets and adversarial rewards
• Stochastic bandits with arm-dependent delays
• Near-linear time Gaussian process optimization with adaptive batching and resparsification
• Fixed-confidence guarantees for Bayesian best-arm identification
• Multiagent evaluation under incomplete information
• Exact sampling of determinantal point processes with sublinear time preprocessing
• Exploiting structure of uncertainty for efficient matroid semi-bandits
• DPPy: Sampling determinantal point processes with Python
• Rotting bandits are not harder than stochastic ones
• Finding the bandit in a graph: Sequential search-and-stop
• Optimistic optimization of a Brownian
• Second-order kernel online convex optimization with adaptive sketching
• Zonotope hit-and-run for efficient sampling from projection DPPs
• Distributed adaptive sampling for kernel matrix approximation
• Simple regret for infinitely many armed bandits
• Cheap Bandits
• Geometric entropic exploration,
• On the approximation relationship between optimizing ratio of submodular (RS) and difference of submodular (DS) functions,
• Learning to Act Greedily: Polymatroid Semi-Bandits

Citation

If you use any of these papers, please cite the original work.

Contact

• Website: https://researchers.mila.quebec/en/profile/michal-valko
• HAL: https://hal.science/search/index/?q=authIdHal_s:michal