Berkeley AI Research
Apr 20
Gradient-based Planning for World Models at Longer Horizons
grasp-results-table table { font-size: 0. 875rem; line-height: 1.
RL without TD learning
In this post, I’ll introduce a reinforcement learning (RL) algorithm based on an “alternative” paradigm: divide and conquer . Unlike traditional methods, this algorithm is not based on temporal difference (TD) learning (which has scalability challenges ), and scales well to long-horizon tasks.
Related stories
arXiv AI
Jun 30
Exploration and Online Transfer with Behavioral Foundation Models
arXiv:2606. 29980v1 Announce Type: new Abstract: Zero-shot Transfer in Reinforcement Learning (RL) aims to train an agent that can generate optimal policies for any reward function, without additional learning at transfer time, while training only on reward-free trajectories.
arXiv Machine Learning
Jun 16
Fast Non-Episodic Finite-Horizon RL with K-Step Lookahead Thresholding
arXiv:2602. 00781v2 Announce Type: replace Abstract: Online reinforcement learning in non-episodic, finite-horizon MDPs remains underexplored and is challenged by the need to estimate returns to a fixed terminal time.
Berkeley AI Research
Sep 1, 2025
What exactly does word2vec learn?
What exactly does word2vec learn, and how? Answering this question amounts to understanding representation learning in a minimal yet interesting language modeling task.
arXiv AI
6d ago
Reward-rate Policy Gradient for Efficient Machine Learning Engineering Agents
arXiv:2609.36393v1 Announce Type: cross Abstract: Traditional reinforcement learning (RL) techniques focus on maximizing expected cumulative reward, where each action assumes to take a constant unit...
arXiv AI
Sep 25
PoEM: Predicting RL Outcomes from Existing Policies
PoEM predicts reinforcement learning outcomes for a new reward function using models already trained on other rewards. If the new reward is a linear combination of existing ones, the new policy’s log-space representation can be expressed as a linear combination of existing log-policies. Even when rewards are not linearly related, log-policies often span a low‑rank subspace, allowing the weighting coefficients to be estimated from reward or basis policy outputs, enabling policy approximation without additional RL training.
Google AI Blog
Mar 14, 2024
Cappy: Outperforming and boosting large multi-task language models with a small scorer
Posted by Yun Zhu and Lijuan Liu, Software Engineers, Google Research Large language model (LLM) advancements have led to a new paradigm that unifies various natural language processing (NLP) tasks within an instruction-following framework. This paradigm is exemplified by recent multi-task LLMs, such as T0 , FLAN , and OPT-IML .
arXiv AI
Jun 18
Reinforcement Learning Foundation Models Should Already Be A Thing
arXiv:2606. 18812v1 Announce Type: cross Abstract: Foundation models for language and vision are powered by internet-scale data, while structured domains (tabular prediction, time-series forecasting, graph learning, reinforcement learning) are not.
arXiv Machine Learning
4d ago
Role-Adaptive Policy Optimization for Offline Reinforcement Learning
arXiv:2609.40149v1 Announce Type: new Abstract: Policy regularization in offline reinforcement learning balances policy improvement against reliance on uncertain value estimates. This balance can dif...
Berkeley AI Research
May 8
Adaptive Parallel Reasoning: The Next Paradigm in Efficient Inference Scaling
apr-fig { text-align: center; margin: 1. 35em 0; line-height: 1.
arXiv AI
4d ago
ALER: Adaptive Learnable Experience Rewriting for Reinforcement Learning
arXiv:2610.00592v1 Announce Type: cross Abstract: In partially observable reinforcement learning (RL), a later observation can make stored information obsolete or change what it implies for the next...
Hugging Face Trending Papers
Jun 17
Reinforcement Learning Foundation Models Should Already Be A Thing
Foundation models for language and vision are powered by internet-scale data, while structured domains (tabular prediction, time-series forecasting, graph learning, reinforcement learning) are not. The substitute is synthetic data, which shifts the burden from collection to prior design.