Stagewise Reinforcement Learning and the Geometry of the Regret Landscape
cs.LG
Submitted: 2026-01-12
Updated: 2026-11-06
Terminology
Sources
- Structural Inference: Interpreting Small Language Models with Susceptibilities
- Dynamics of Transient Structure in In-Context Linear Regression Transformers
- Dynamical versus Bayesian Phase Transitions in a Toy Model of Superposition
- Why neural networks find simple solutions: the many regularizers of geometric complexity
- IMPALA: Scalable Distributed Deep-RL with Importance Weighted Actor-Learner Architectures
- The Statistical Complexity of Interactive Decision Making
- Scaling Laws for Reward Model Overoptimization
- Cooperative Inverse Reinforcement Learning
- From Global to Local: A Scalable Benchmark for Local Posterior Sampling
- On Large-Batch Training for Deep Learning: Generalization Gap and Sharp Minima
- Bayesian Influence Functions for Hessian-Free Data Attribution
- SimBa: Simplicity Bias for Scaling Up Parameters in Deep Reinforcement Learning
- Reinforcement Learning and Control as Probabilistic Inference: Tutorial and Review
- Preconditioned Stochastic Gradient Langevin Dynamics for Deep Neural Networks
- Understanding and Controlling a Maze-Solving Policy Network
- Simplicity Bias in 1-Hidden Layer Neural Networks
- You Are What You Eat -- AI Alignment Requires Understanding How Data Shapes Structure and Generalisation
- Deep learning generalizes because the parameter-function map is biased towards simple functions
- Colour versus Shape Goal Misgeneralization in Reinforcement Learning: A Case Study
- The Transient Nature of Emergent In-Context Learning in Transformers
Related papers
- FlexMoE: Learning One-for-All Multi-Budget Expert Pruning for Flexible MoE Deployment
- OptiMind: Improving Small Language Models for Optimization Modeling through Formulation Guidance and Multi-Turn Learning
- Staying in the Sweet Spot: Boost Reasoning Evolution via Responsive Rollout Difficulty Manipulation
- Examining the Impact of Data and Training Methods on Emergent and Subliminal Misalignment
- Rethinking Black-Box On-Policy Distillation: A Rubric-based Analysis
- (Mis)generalization of Helpful-Only Fine-tuning