Conditions on Preference Relations that Guarantee the Existence of Optimal Policies

• URL: http://arxiv.org/abs/2311.01990v2
• Date: Wed, 27 Mar 2024 22:03:46 GMT
• Title: Conditions on Preference Relations that Guarantee the Existence of Optimal Policies
• Authors: Jonathan Colaço Carr, Prakash Panangaden, Doina Precup,
• Abstract summary: We introduce a new framework for analyzing LfPF problems in partially-observable, non-Markovian environments. We show that a decision-making problem can have optimal policies even when no reward function can express the learning goal.
• Score: 38.17324903156351
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Learning from Preferential Feedback (LfPF) plays an essential role in training Large Language Models, as well as certain types of interactive learning agents. However, a substantial gap exists between the theory and application of LfPF algorithms. Current results guaranteeing the existence of optimal policies in LfPF problems assume that both the preferences and transition dynamics are determined by a Markov Decision Process. We introduce the Direct Preference Process, a new framework for analyzing LfPF problems in partially-observable, non-Markovian environments. Within this framework, we establish conditions that guarantee the existence of optimal policies by considering the ordinal structure of the preferences. We show that a decision-making problem can have optimal policies -- that are characterized by recursive optimality equations -- even when no reward function can express the learning goal. These findings underline the need to explore preference-based learning strategies which do not assume that preferences are generated by reward.

Related papers

• Optimization Solution Functions as Deterministic Policies for Offline Reinforcement Learning [7.07623669995408]
  We propose an implicit actor-critic (iAC) framework that employs optimization solution functions as a deterministic policy (actor) and a monotone function over the optimal value of optimization as a critic. We show that the learned policies are robust to the suboptimality of the learned actor parameters via the exponentially decaying sensitivity (EDS) property. We validate the proposed framework on two real-world applications and show a significant improvement over state-of-the-art (SOTA) offline RL methods.
  arXiv  Detail & Related papers  (2024-08-27T19:04:32Z)
• Provably Mitigating Overoptimization in RLHF: Your SFT Loss is Implicitly an Adversarial Regularizer [52.09480867526656]
  We identify the source of misalignment as a form of distributional shift and uncertainty in learning human preferences. To mitigate overoptimization, we first propose a theoretical algorithm that chooses the best policy for an adversarially chosen reward model. Using the equivalence between reward models and the corresponding optimal policy, the algorithm features a simple objective that combines a preference optimization loss and a supervised learning loss.
  arXiv  Detail & Related papers  (2024-05-26T05:38:50Z)
• Towards Efficient Exact Optimization of Language Model Alignment [93.39181634597877]
  Direct preference optimization (DPO) was proposed to directly optimize the policy from preference data. We show that DPO derived based on the optimal solution of problem leads to a compromised mean-seeking approximation of the optimal solution in practice. We propose efficient exact optimization (EXO) of the alignment objective.
  arXiv  Detail & Related papers  (2024-02-01T18:51:54Z)
• Nash Learning from Human Feedback [86.09617990412941]
  We introduce an alternative pipeline for the fine-tuning of large language models using pairwise human feedback. We term this approach Nash learning from human feedback (NLHF) We present a novel algorithmic solution, Nash-MD, founded on the principles of mirror descent.
  arXiv  Detail & Related papers  (2023-12-01T19:26:23Z)
• PARL: A Unified Framework for Policy Alignment in Reinforcement Learning from Human Feedback [106.63518036538163]
  We present a novel unified bilevel optimization-based framework, textsfPARL, formulated to address the recently highlighted critical issue of policy alignment in reinforcement learning. Our framework addressed these concerns by explicitly parameterizing the distribution of the upper alignment objective (reward design) by the lower optimal variable. Our empirical results substantiate that the proposed textsfPARL can address the alignment concerns in RL by showing significant improvements.
  arXiv  Detail & Related papers  (2023-08-03T18:03:44Z)
• Policy Gradient Algorithms Implicitly Optimize by Continuation [7.351769270728942]
  We argue that exploration in policy-gradient algorithms consists in a continuation of the return of the policy at hand, and that policies should be history-dependent rather than to maximize the return.
  arXiv  Detail & Related papers  (2023-05-11T14:50:20Z)
• Sample Efficient Model-free Reinforcement Learning from LTL Specifications with Optimality Guarantees [17.69385864791265]
  We present a model-free Reinforcement Learning (RL) approach that efficiently learns an optimal policy for an unknown system. We also provide improved theoretical results on choosing the key parameters to ensure optimality.
  arXiv  Detail & Related papers  (2023-05-02T12:57:05Z)
• A Parametric Class of Approximate Gradient Updates for Policy Optimization [47.69337420768319]
  We develop a unified perspective that re-expresses the underlying updates in terms of a limited choice of gradient form and scaling function. We obtain novel yet well motivated updates that generalize existing algorithms in a way that can deliver benefits both in terms of convergence speed and final result quality.
  arXiv  Detail & Related papers  (2022-06-17T01:28:38Z)
• Near Optimal Policy Optimization via REPS [33.992374484681704]
  emphrelative entropy policy search (REPS) has demonstrated successful policy learning on a number of simulated and real-world robotic domains. There exist no guarantees on REPS's performance when using gradient-based solvers. We introduce a technique that uses emphgenerative access to the underlying decision process to compute parameter updates that maintain favorable convergence to the optimal regularized policy.
  arXiv  Detail & Related papers  (2021-03-17T16:22:59Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.