New method learns from either positive or negative feedback alone.
problem Limited applicability of existing preference optimization methods in scenarios with only unpaired feedback.
method Decouples learning from positive and negative feedback, using expectation-maximization (EM) to optimize probability of positive outcomes and explicitly incorporate negative examples.
result Stable learning from negative feedback alone demonstrated.
Study evaluates new models using human feedback from another model.
problem Evaluate a new model using human feedback collected for another model.
method Formalize problem, propose model-based and model-free estimators, analyze unbiasedness, and empirically evaluate.
result Proposed estimators can predict absolute values, rank, and optimize evaluated policies.
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.
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.
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,…
Classifier learns to ignore unreliable feedback from end users.
problem Improving classifier performance by filtering unreliable feedback.
method Modeling end users as autonomous agents, periodically retraining classifier with filtered feedback.
result Classifier can identify and filter out unreliable feedback, improving performance.
Counterfactual learning from human bandit feedback describes a scenario where user feedback on the quality of outputs of a historic system is logged and used to improve a target system. We show how to apply this learning framework to neural semantic parsing. From a machine learning perspective, the key challenge lies i…
User preferences for items can be inferred from either explicit feedback, such as item ratings, or implicit feedback, such as rental histories. Research in collaborative filtering has concentrated on explicit feedback, resulting in the development of accurate and scalable models. However, since explicit feedback is oft…
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.
The paper proposes a new method for learning reward models from ordinal feedback, improving upon binary feedback.
problem Learning reward models from human preferences using binary feedback discards useful samples and loses fine-grained information.
method The paper introduces a framework for learning reward models under ordinal feedback, generalizing the Bradley-Terry model.
result Ordinal feedback reduces the Rademacher complexity compared to binary feedback, leading to better reward learning.
Paper introduces robust learning from feature feedback, even with annotator errors.
problem Learning from human feedback on discriminative features, especially when annotators make mistakes.
method Develops a robust framework for learning with imperfect feedback, deriving regret bounds in adversarial and stochastic settings.
result Regret bounds independent of feature number, showing robust learning is not reducible to non-robust settings.
New algorithm for active learning from feedback coding.
problem Efficiently selecting examples for labeling in active machine learning.
method Formalized structural similarities between active learning and feedback channel coding, developed an optimal transport-based algorithm called Approximate Posterior Matching (APM).
result Learning performance comparable to existing methods at reduced computational cost.
Interactive Machine Learning is concerned with creating systems that operate in environments alongside humans to achieve a task. A typical use is to extend or amplify the capabilities of a human in cognitive or physical ways, requiring the machine to adapt to the users' intentions and preferences. Often, this takes the…
Estimates mode from partial feedback, improving AI learning pipelines.
problem Estimating the mode of a distribution with partial feedback.
method Entropy coding, coarse sufficient statistics, bandit algorithms.
result Statistically and computationally efficient solution to mode estimation.
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.
Study analyzes feedback complexity for sparse feature retrieval in deep networks.
problem Learning sparse superposed features with feedback.
method Analysis of feedback complexity in sparse settings, including triplet comparisons.
result Establishes tight bounds and strong upper bounds for feature recovery.
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.
We study an online decision making problem where on each round a learner chooses a list of items based on some side information, receives a scalar feedback value for each individual item, and a reward that is linearly related to this feedback. These problems, known as contextual semibandits, arise in crowdsourcing, rec…
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.
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.
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.
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.
Paper fine-tunes LLMs using user edits, unifying preference, supervision, and reward feedback.
problem Adapting LLMs to user preferences and feedback types.
method Derives bounds for learning algorithms from user edits, proposes an ensembling procedure.
result Ensembling procedure outperforms individual feedback methods and robustly adapts to different user-edit distributions.
Learning from human feedback is a viable alternative to control design that does not require modelling or control expertise. Particularly, learning from corrective advice garners advantages over evaluative feedback as it is a more intuitive and scalable format. The current state-of-the-art in this field, COACH, has pro…
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.
Study how communication and feedback graphs affect learning outcomes.
problem Understanding the impact of feedback graphs on cooperative online learning.
method Analyzed network regret in terms of the independence number of the strong product of communication and feedback graphs.
result Proved bounds for network regret and demonstrated the non-improvable nature of positive results in pathological cases.
Algorithm learns fair division from noisy feedback in uncertain markets.
problem Learning fair division in uncertain markets with noisy feedback.
method Wrapper algorithms using dual averaging to learn item and agent values from bandit feedback.
result Asymptotically achieves optimal Nash social welfare in linear Fisher markets.
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.
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.
Traditional collaborative filtering (CF) based recommender systems tend to perform poorly when the user-item interactions/ratings are highly scarce. To address this, we propose a learning framework that improves collaborative filtering with a synthetic feedback loop (CF-SFL) to simulate the user feedback. The proposed …
First sample-efficient algorithm for learning EFCE in bandit feedback settings.
problem Learning EFCE in bandit feedback settings for IIEFGs.
method Proposed K-EFCE and uncoupled no-regret algorithm with wide-range regret minimization. result First sample-efficient algorithm for learning EFCE from bandit 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…
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.
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.
Graph-based approach repairs programs from diagnostic feedback.
problem Learning to repair programs from limited labeled data and compiler error messages.
method Introduces program-feedback graph and graph neural network for reasoning, and self-supervised learning with unlabeled programs.
result DrRepair significantly outperforms prior work, achieving high repair rates.
Paper addresses generalization error bounds for learning with censored feedback.
problem Impact of censored feedback on generalization error bounds.
method Derives an extension of DKW inequality for non-IID data due to censored feedback and uses it to bound generalization error.
result Existing generalization error bounds fail to account for censored feedback, necessitating new bounds.
Novel algorithms for online learning with uncertain feedback graphs reduce regret.
problem Uncertainty in feedback graphs hinders traditional online learning approaches.
method Developed novel online learning algorithms to handle uncertain feedback graphs.
result Proved sublinear regret under mild conditions for the proposed algorithms.
This work compares human feedback methods for reward learning in bandits.
problem Understanding how human feedback affects the performance of reward learning methods.
method Theoretical comparison of human feedback approaches in offline contextual bandits.
result Human bias and uncertainty in feedback modeling impact the theoretical guarantees of reward learning methods.
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 study an online learning framework introduced by Mannor and Shamir (2011) in which the feedback is specified by a graph, in a setting where the graph may vary from round to round and is \emph{never fully revealed} to the learner. We show a large gap between the adversarial and the stochastic cases. In the adversaria…
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 consider combinatorial online learning with subset choices when only relative feedback information from subsets is available, instead of bandit or semi-bandit feedback which is absolute. Specifically, we study two regret minimisation problems over subsets of a finite ground set [n], with subset-wise relative prefe…
Interactive weak supervision learns useful heuristics from user feedback.
problem Creating useful heuristics for large labeled datasets is tedious and subjective.
method Develops an interactive framework for learning heuristics from user feedback.
result Only a few feedback iterations are needed to train models without ground truth labels.
DOPL learns from preference feedback to solve RMAB problems.
problem Learning optimal decisions in RMAB with limited reward information.
method Direct online preference learning (DOPL) for Pref-RMAB.
result DOPL achieves sublinear regret for RMAB with preference feedback.
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.
Artificial neural networks are most commonly trained with the back-propagation algorithm, where the gradient for learning is provided by back-propagating the error, layer by layer, from the output layer to the hidden layers. A recently discovered method called feedback-alignment shows that the weights used for propagat…
Sayer uses implicit feedback to optimize system policies.
problem Leveraging implicit feedback to improve system policies is difficult due to bias and incompleteness.
method Sayer combines randomized exploration and unbiased counterfactual estimators to evaluate and train new policies using implicit feedback.
result Sayer can accurately evaluate and train new policies that outperform existing ones.
The problem of Reinforcement Learning (RL) in an unknown nonlinear dynamical system is equivalent to the search for an optimal feedback law utilizing the simulations/ rollouts of the dynamical system. Most RL techniques search over a complex global nonlinear feedback parametrization making them suffer from high trainin…