Global ETD Search

Search theses and dissertations gathered from participating repositories worldwide. Every result links back to the library that holds it. No account is needed.

Results

Showing 1 to 20 of 33 for “"RLHF"”.

  1. NLP Analysis in the Intersection of RLHF and NER

    … reinforcement learning with human feedback(RLHF) into the NER framework. By using an actor-critic Re- inforcement Learning paradigm, the model continuously improves recognition through iterative feedback and reward-based mechanisms. The proposed system employs GloVe embeddings to capture …

    ttu Repository record for NLP Analysis in the Intersection of RLHF and NER (opens in a new tab)

  2. Algorithmic Foundations for Safe and Efficient Reinforcement Learning from Human Feedback

    … Reinforcement Learning from Human Feedback (RLHF) allows RL agents to learn from human-provided data, such as evaluations or rankings of trajectories. In many applications, human feedback is expensive to collect; therefore, learning robust policies from limited data is crucial. In this …

    ethz Repository record for Algorithmic Foundations for Safe and Efficient Reinforcement Learning from Human Feedback (opens in a new tab)

  3. Designing, learning, and utilizing dense reward for generative AI alignment

    … for reinforcement-learning-from-human-feedback (RLHF) training of text and text-to-image generation models. We start from learning better dynamics and reward models for policy learning in offline model-based reinforcement learning (offline MBRL), from the angle of mitigating the mismatched …

    texas Repository record for Designing, learning, and utilizing dense reward for generative AI alignment (opens in a new tab)

  4. On the current landscape of language model reward modeling for alignment

    Reinforcement Learning with Human Feedback (RLHF) has become a central approach for aligning large language models (LLMs) with human preferences. Despite its success, a significant gap remains between human judgment and LLM behavior, particularly in automated evaluation settings and in capturing …

    umn Repository record for On the current landscape of language model reward modeling for alignment (opens in a new tab)

  5. On the current landscape of language model reward modeling for alignment

    Reinforcement Learning with Human Feedback (RLHF) has become a central approach for aligning large language models (LLMs) with human preferences. Despite its success, a significant gap remains between human judgment and LLM behavior, particularly in automated evaluation settings and in capturing …

    umn Repository record for On the current landscape of language model reward modeling for alignment (opens in a new tab)

  6. Aligning Language Models Using Multi-Objective Deep Reinforcement Learning

    … is reinforcement learning from human feedback (RLHF). RLHF aims to optimize one objective based on human preferences. However, the cost of high-quality human feedback is enormous. Having all human annotators consistent in their opinions on desirable behaviors is also challenging. LLM alignment …

    brock Repository record for Aligning Language Models Using Multi-Objective Deep Reinforcement Learning (opens in a new tab)

  7. Deep-Reinforcement Learning Agents in Autonomous Directional Drilling and Geosteering

    … Reinforcement Learning from Human Feedback (RLHF) was integrated to align automated decisions with expert knowledge and field practices. This framework provides scope for embedding engineer and geologist feedback directly into the reward model. A preference-based neural network is modelled to …

    tamu Repository record for Deep-Reinforcement Learning Agents in Autonomous Directional Drilling and Geosteering (opens in a new tab)

  8. Deep-Reinforcement Learning Agents in Autonomous Directional Drilling and Geosteering

    … Reinforcement Learning from Human Feedback (RLHF) was integrated to align automated decisions with expert knowledge and field practices. This framework provides scope for embedding engineer and geologist feedback directly into the reward model. A preference-based neural network is modelled to …

    tamu Repository record for Deep-Reinforcement Learning Agents in Autonomous Directional Drilling and Geosteering (opens in a new tab)

  9. Deep-Reinforcement Learning Agents in Autonomous Directional Drilling and Geosteering

    … Reinforcement Learning from Human Feedback (RLHF) was integrated to align automated decisions with expert knowledge and field practices. This framework provides scope for embedding engineer and geologist feedback directly into the reward model. A preference-based neural network is modelled to …

    tamu Repository record for Deep-Reinforcement Learning Agents in Autonomous Directional Drilling and Geosteering (opens in a new tab)

  10. Deep-Reinforcement Learning Agents in Autonomous Directional Drilling and Geosteering

    … Reinforcement Learning from Human Feedback (RLHF) was integrated to align automated decisions with expert knowledge and field practices. This framework provides scope for embedding engineer and geologist feedback directly into the reward model. A preference-based neural network is modelled to …

    tamu Repository record for Deep-Reinforcement Learning Agents in Autonomous Directional Drilling and Geosteering (opens in a new tab)

  11. Investigating and Tackling Bias and Hackings in AI Alignment

    … is Reinforcement Learning from Human Feedback~(RLHF), which trains models by optimizing them against a reward signal derived from human preferences. While effective, this paradigm is susceptible to failure modes where models learn to maximize their reward score without genuinely adhering to the …

    maryland Repository record for Investigating and Tackling Bias and Hackings in AI Alignment (opens in a new tab)

  12. Across the Reinforcement Learning and Large Language Model Spectrum

    … reinforcement learning from human feedback (RLHF), including sampler-dependent convergence for online direct preference optimization (DPO) and last-iterate convergence guarantees for Nash learning under non-transitive preferences. The final part turns to RL for LLMs, presenting …

    washington Repository record for Across the Reinforcement Learning and Large Language Model Spectrum (opens in a new tab)

  13. Exploring Optimization Strategies for Multi-Objective Reward Modeling

    … has become a standard for this purpose. In the RLHF, reward models play an impotent role. It was used to approximate human judgment, the reward model will provide a signal during the reinforcement learning stage. However, most existing reward models rely on a single scalar reward, which may not …

    brock Repository record for Exploring Optimization Strategies for Multi-Objective Reward Modeling (opens in a new tab)

  14. PRINCIPLED FRAMEWORKS FOR AI ALIGNMENT: FROM POST-TRAINING TO INFERENCE

    … with reinforcement learning from human feedback (RLHF) - its failure to capture the entanglement between reward learning and policy optimization which leads to a distribution shift and suboptimal alignment. To mitigate this, we propose a novel and efficient bilevel alignment framework that …

    maryland Repository record for PRINCIPLED FRAMEWORKS FOR AI ALIGNMENT: FROM POST-TRAINING TO INFERENCE (opens in a new tab)

  15. Co-design of communication, computing and control for real-time interactions in industrial cyber-physical systems

    … Reinforcement Learning from Human Feedback (RLHF) into the co-design loop. By modeling implicit human responses as latent reward signals, the framework learns control strategies that optimize both task efficiency and perceived comfort. We validate our framework using a UR3e robotic arm for …

    glasgow Repository record for Co-design of communication, computing and control for real-time interactions in industrial cyber-physical systems (opens in a new tab) Download PDF for Co-design of communication, computing and control for real-time interactions in industrial cyber-physical systems (opens in a new tab)

  16. Aligning language models to professional domains using preference training

    … as reinforcement learning from human feedback (RLHF) or direct preference optimization (DPO), can significantly improve the alignment of models with user intent and linguistic requirements. By using these training methods, smaller models have also been shown to be able to produce solutions that …

    reykjavik Repository record for Aligning language models to professional domains using preference training (opens in a new tab)

  17. Enhancing Empathy in Conversational AI with Interpretable Lexicons and Reinforcement Learning from AI Feedback (RLAIF)

    … Reinforcement Learning from Human Feedback (RLHF) has shown promise in aligning model behavior, but human feedback is costly, subjective, and does not scale. In order to address these issues, this thesis addresses the above two interconnected challenges through three contributions. First, we …

    texas-state Repository record for Enhancing Empathy in Conversational AI with Interpretable Lexicons and Reinforcement Learning from AI Feedback (RLAIF) (opens in a new tab)

  18. Human Comfort Modeling, Measurement, and Improvement in Human–Robot Collaboration

    … (LLMs) for adaptive planning, utilizing an RLHF-inspired approach to align task planning strategies with human comfort preferences; and (3) enhancing robot intention communication through augmented reality (AR), ensuring that humans can intuitively perceive robot intentions and actions, …

    clemson Repository record for Human Comfort Modeling, Measurement, and Improvement in Human–Robot Collaboration (opens in a new tab)

  19. On Safety in Sequential Decision Learning

    … Reinforcement Learning from Human Feedback (RA-RLHF), which incorporates Conditional Value at Risk principles during model fine-tuning. This approach effectively reduces toxic content generation while maintaining model performance on general tasks, contributing to safer AI deployment. Addressing …

    tamu Repository record for On Safety in Sequential Decision Learning (opens in a new tab)

  20. Improving the Quality, Efficiency and Understanding of Generative Models

    … multimodal models. ShortCoTI introduces a RLHF (Reinforcement Learning from Human Feedback)-based optimization that reduces redundant reasoning steps in autoregressive image generation while maintaining or improving output quality. Together, these contributions advance the development of …

    cornell Repository record for Improving the Quality, Efficiency and Understanding of Generative Models (opens in a new tab) Embargoed until 2028-01-08

Page 1 of 2