LoCo-RLHF models diverse human feedback with contextual information.
problem Heterogeneous human feedback from diverse contexts and preferences.
method Low-rank contextual preference model, PRS policy.
result LoCo-RLHF achieves tighter sub-optimality gap than existing methods.
The abstract explores connections between reinforcement learning, scaling, and diffusion.
problem Aligning reinforcement learning with human feedback and scaling techniques.
method Clarifying connections between reinforcement learning, scaling, and diffusion.
result Introducing a resampling approach for alignment and reward-directed diffusion models.
Study preference-based reinforcement learning in episodic kernel MDPs.
problem Learning from episodic human preferences in reinforcement learning.
method Developed preference-based value estimation and confidence sets for kernel-based MDPs.
result Proved high-probability regret bounds that converge to optimal policy value.
New STDP rule for spiking neurons solves discrete action reinforcement learning tasks.
problem Applying standard STDP to discrete action reinforcement learning tasks.
method Feedback-modulated TD-STDP learning rule for spiking neuron networks.
result Feedback modulation improves credit assignment in reinforcement learning.
RLHF uses human feedback to train AI models, posing statistical challenges.
problem Aligning AI models with human preferences using noisy, subjective feedback.
method Supervised fine-tuning, reward modeling, policy optimization, statistical ideas.
result Statistical methods for reward function learning and policy optimization.
Unified LP framework for offline reward learning from human demonstrations and feedback.
problem Reward learning from human demonstrations and feedback with robustness and sample efficiency.
method A novel linear programming framework for offline reward learning.
result Unified LP framework achieves better performance compared to MLE.
Active learning framework for optimizing human preferences in reinforcement learning.
problem Selecting most informative feedback for training models of human preferences.
method Proposes an active learning framework to collect preferential feedback online or offline.
result Errors in DPO logit estimates diminish with more feedback.
Advocates a local feedback approach for RL in unknown systems.
problem Finding optimal feedback laws in unknown nonlinear dynamical systems.
method Searches over a local feedback representation consisting of an open-loop sequence and an optimal linear feedback law.
result Results in highly efficient training and superior performance compared to global methods.
Safe RL with binary feedback using SABRE algorithm.
problem Safe reinforcement learning with binary safety feedback.
method SABRE algorithm, combining active learning and reinforcement learning.
result Provable safe policy with high probability, no unsafe actions during training.
Dual active learning improves RLHF by selecting optimal conversations and teachers.
problem Efficiently aligning LLMs with human preferences using RLHF from feedback.
method Offline RL for conversation and teacher selection, dual active reward learning, pessimistic RL.
result The proposed algorithm achieves minimal generalized variance and outperforms state-of-the-arts.
Reinforcement learning with trajectory feedback instead of state-action rewards.
problem Frequent feedback not available in practice.
method Extended reinforcement learning algorithms using trajectory feedback for known and unknown transition models.
result Hybrid optimistic-Thompson Sampling algorithm for unknown transition models.
Study of reinforcement learning with additional feedback observations.
problem Episodic reinforcement learning in Markov decision processes with feedback observations.
method Formalization of feedback graph, model-based algorithms leveraging feedback, regret bound analysis.
result Regret bound depends only on the size of the maximum acyclic subgraph of the feedback graph.
We present a study on reinforcement learning (RL) from human bandit feedback for sequence-to-sequence learning, exemplified by the task of bandit neural machine translation (NMT). We investigate the reliability of human bandit feedback, and analyze the influence of reliability on the learnability of a reward estimator,…
RLHF fails when humans only partially observe, leading to inflated or overjustified feedback.
problem Failure of reinforcement learning from human feedback in partially observable environments.
method Formal definition of failure cases, modeling human as Boltzmann rational, analyzing information provided by feedback.
result RLHF can deceptively inflate or overjustify feedback when humans have partial observations.
Study shows reinforcement learning is possible with once-per-episode feedback.
problem Challenges of reinforcement learning with sparse feedback.
method Introduced a model where trajectory labels are generated by an unknown parametric model and developed an algorithm for sublinear regret.
result Achieved sublinear regret with a statistically and computationally efficient algorithm.
Generative model solves financial market equilibria with stable reinforcement learning.
problem Financial market equilibria under realistic frictions and multiple agents.
method Generative adversarial reinforcement learning with decoupling feedback.
result Algorithm learns and predicts asset returns and volatilities.
Meta-algorithm for efficient reinforcement learning from human preferences.
problem Learning from human preference comparisons in Markov decision processes.
method Randomized exploration and experimental design for batch comparison queries.
result Meta-algorithm achieves both regret and last-iterate guarantees with minimal preference queries.
Paper analyzes online reinforcement learning with outcome-based feedback, providing efficient algorithms and fundamental limits.
problem Assigning credit to actions in reinforcement learning with only endpoint rewards.
method Develops a provably sample-efficient algorithm for online reinforcement learning with general function approximation.
result Achieves O ( C m c o v H 3 / ε 2 ) O(C_{
m cov} H^3/ε^2) O ( C m co v H 3 / ε 2 ) sample complexity, characterizing statistical separation between outcome-based and per-step rewards. This paper argues for more realistic human models in RL.
problem Current RL models oversimplify human feedback, ignoring personal, contextual, and dynamic aspects.
method Calls for interdisciplinary research on human feedback in RL.
result Realistic human models are needed for robust human-in-the-loop RL systems.
Develops a new model for RLHF accounting for partially observed states and intermediate feedback.
problem Lack of models for partially observed states and intermediate feedback in RLHF.
method PORRL model with cardinal and dueling feedback methods.
result Demonstrates improved learning and alignment with new model-based and model-free methods.
Efficiently identifies good policies by choosing contexts for human feedback.
problem Efficiently identifying good policies in applications with high feedback costs.
method Introduces offline contextual dueling bandit setting and an upper-confidence-bound style algorithm.
result Proves a regret bound and shows superior performance over uniformly sampled contexts.
Privacy-preserving reinforcement learning from human feedback using decoupled reward modeling.
problem Training large language models with sensitive user information while preserving privacy.
method Proposes a privacy-preserving framework that imposes differential privacy on reward learning only.
result Privacy contributes an additional additive term to the suboptimality gap, and the upper bound is rate-optimal up to logarithmic factors.
PILAF optimizes reward models from human feedback for better policy alignment.
problem Creating accurate reward models from human feedback for policy optimization.
method Policy-Interpolated Learning for Aligned Feedback (PILAF) that explicitly aligns preference learning with maximizing underlying oracle reward.
result PILAF is optimal from both optimization and statistical perspectives, demonstrating strong performance in RLHF settings.
Paper tackles combinatorial reinforcement learning with preference feedback.
problem Modeling long-term user engagement in scenarios like recommender systems and online advertising.
method Assumes a contextual MNL preference model with linear mean utilities and approximates item values. Proposes MNL-VQL algorithm.
result Achieves nearly minimax-optimal regret for linear MDPs with preference feedback.
Recommender systems play a crucial role in mitigating the problem of information overload by suggesting users' personalized items or services. The vast majority of traditional recommender systems consider the recommendation procedure as a static process and make recommendations following a fixed strategy. In this paper…
ADPO optimizes relative advantage in reinforcement learning from human feedback.
problem Optimizing policy alignment in reinforcement learning from human preferences.
method ADPO explicitly parameterizes the optimal structure through anchored logits, decoupling response quality from prior popularity.
result Empirically, ADPO achieves state-of-the-art performance on reasoning tasks, outperforming GRPO by 30.9 percent.
New framework solves dynamic bilevel optimization problems in reinforcement learning.
problem Dynamic objective functions in reinforcement learning and human feedback.
method Principled penalty-based methods for bilevel reinforcement learning.
result Demonstrated effectiveness of penalty-based algorithms in simulations.
Variational Proximal Policy Optimization improves reinforcement learning from human feedback.
problem Policy mode collapse and brittle exploration loops in reinforcement learning.
method Particle-based variational inference framework with Mixture-of-Experts architecture.
result Significant improvements in complex reasoning benchmarks.
Study shows how to control jump-diffusion processes with stable feedback controls in reinforcement learning.
problem Control jump-diffusion processes with unknown coefficients in reinforcement learning.
method Lipschitz continuous optimal feedback controls, stability analysis of forward-backward SDEs, least-squares algorithm.
result Achieves O ( N ln N ) O(\sqrt{N\ln N}) O ( N ln N ) regret for linear-convex learning problems with jumps. The paper analyzes RLHF with human feedback and provides convergence results for MLE and pessimistic MLE.
problem Improving RLHF with human feedback from pairwise or K K K -wise comparisons. method Theoretical framework for RLHF with convergence analysis of MLE and pessimistic MLE.
result MLE fails but pessimistic MLE provides improved policies under certain coverage assumptions.
CausalRM models rewards from user feedback, overcoming noise and bias.
problem Aligning language models with user preferences from noisy, biased feedback.
method Causal-theoretic reward modeling framework addressing noise and bias in observational feedback.
result CausalRM learns accurate reward signals from noisy and biased observational feedback.
Improves reinforcement learning for complex tasks with sparse feedback.
problem Learning optimal policies from sparse feedback is challenging.
method Three algorithms based on Hindsight Experience Replay (HER) to improve performances.
result Vast improvement in final success rate and sample efficiency.
Industrial recommender systems deal with extremely large action spaces -- many millions of items to recommend. Moreover, they need to serve billions of users, who are unique at any point in time, making a complex user state space. Luckily, huge quantities of logged implicit feedback (e.g., user clicks, dwell time) are …
Proposes a new theoretical framework for PbRL that requires less human feedback.
problem Lack of theoretical work capturing practical PbRL frameworks.
method Introduces a reward-agnostic PbRL framework that acquires exploratory trajectories before human feedback.
result Demonstrates improved sample complexity for learning optimal policies in linear and low-rank MDPs.
This thesis tackles learning reward functions from human comparative feedback.
problem Designing reward functions for complex tasks is challenging and humans often provide suboptimal demonstrations.
method Proposes learning reward functions from comparative feedback (pairwise comparisons, best-of-many choices, rankings, scaled comparisons) and active learning techniques.
result Demonstrates the effectiveness of learning reward functions from comparative feedback in various domains.
New RL algorithm optimizes policies with bandit feedback, matching previous bounds.
problem Optimizing policies with unknown transitions and bandit feedback.
method Optimistic Trust Region Policy Optimization (TRPO) algorithm.
result Sub-linear regret bounds for both stochastic and adversarial rewards.
New RLHF framework handles general preference oracles without reward functions.
problem Handling general preference oracles without assuming a reward function.
method Developed a minimax game between two LLMs for RLHF under a general preference oracle, focusing on KL-regularized preference.
result Proposed algorithms for efficient offline and online RLHF learning.
GPG improves RL from feedback with less memory and compute.
problem Improving reinforcement learning from human feedback.
method Group-based Monte Carlo advantage estimator replacing learned value function.
result GPG matches or outperforms PPO on standard benchmarks.
Paper proposes a robust RLHF algorithm for LLMs, improving response preference over baselines.
problem Reward model misspecifications in RLHF for LLMs.
method Proposes a robust algorithm to reduce reward and policy estimator variance, theoretically and empirically validated.
result Consistently outperforms existing methods on LLM benchmark datasets, favoring 77-81% of responses over baselines.
A new approach to fine-tuning LLMs with human feedback.
problem Inability of current reward models to fully represent human preferences.
method Introducing NLHF, a new pipeline for LLM fine-tuning using pairwise human feedback.
result NLHF produces a sequence of policies converging to the regularized Nash equilibrium.
We present an approach to interactive-predictive neural machine translation that attempts to reduce human effort from three directions: Firstly, instead of requiring humans to select, correct, or delete segments, we employ the idea of learning from human reinforcements in form of judgments on the quality of partial tra…
SLHF uses sequential game theory to optimize preferences from human feedback.
problem Optimizing preferences from human feedback in sequential settings.
method SLHF frames the problem as a sequential-move game between Leader and Follower, decomposing the optimization into refinement and adversarial optimization.
result SLHF achieves strong alignment across diverse preference datasets and scales to large models.
New Feedback Transformer architecture improves model performance by exposing past representations to future.
problem Limitations of Transformers in fully exploiting sequential input.
method Proposes Feedback Transformer exposing all past representations to future.
result Demonstrates improved performance with smaller, shallower models.
Enhanced feedback model improves sample-efficiency in POMDPs.
problem Exponential hardness of learning in POMDPs.
method Multiple observations in hindsight feedback model.
result Sample-efficient learning possible for new subclasses of POMDPs.
New Q-learning algorithms reduce regret in inventory control problems.
problem Efficiently learning optimal policies in inventory control problems with limited feedback.
method Proposed Elimination-Based Half-Q-Learning (HQL) and Full-Q-Learning (FQL) algorithms with theoretical regret bounds.
result HQL incurs i l d e O ( H 3 T ) ilde{\mathcal{O}}(H^3\sqrt{ T}) i l d e O ( H 3 T ) regret, FQL incurs i l d e O ( H 2 T ) ilde{\mathcal{O}}(H^2\sqrt{ T}) i l d e O ( H 2 T ) regret, independent of state and action space sizes. New method provides fine-grained feedback on interactive student programs.
problem Time-consuming manual grading of interactive student programs.
method Meta-exploration approach using reinforcement learning.
result 94.3% accuracy in providing fine-grained feedback.
Learning optimal feedback control laws capable of executing optimal trajectories is essential for many robotic applications. Such policies can be learned using reinforcement learning or planned using optimal control. While reinforcement learning is sample inefficient, optimal control only plans an optimal trajectory fr…
Deep Reinforcement Learning has enabled the control of increasingly complex and high-dimensional problems. However, the need of vast amounts of data before reasonable performance is attained prevents its widespread application. We employ binary corrective feedback as a general and intuitive manner to incorporate human …