Multi-head Reward Aggregation Guided by Entropy

• URL: http://arxiv.org/abs/2503.20995v1
• Date: Wed, 26 Mar 2025 21:16:48 GMT
• Title: Multi-head Reward Aggregation Guided by Entropy
• Abstract summary: Rules characterized by high rating entropy are less reliable in identifying responses preferred by humans.<n>We introduce ENCORE, a straightforward entropy-guided approach that composes multi-head rewards by downweighting rules exhibiting high rating entropy.<n>Our proposed approach is training-free, broadly applicable to various datasets, and maintains interpretability.
• Score: 1.7505359879440667
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Aligning large language models (LLMs) with safety guidelines typically involves reinforcement learning from human feedback (RLHF), relying on human-generated preference annotations. However, assigning consistent overall quality ratings is challenging, prompting recent research to shift towards detailed evaluations based on multiple specific safety criteria. This paper uncovers a consistent observation: safety rules characterized by high rating entropy are generally less reliable in identifying responses preferred by humans. Leveraging this finding, we introduce ENCORE, a straightforward entropy-guided approach that composes multi-head rewards by downweighting rules exhibiting high rating entropy. Theoretically, we demonstrate that rules with elevated entropy naturally receive minimal weighting in the Bradley-Terry optimization framework, justifying our entropy-based penalization. Through extensive experiments on RewardBench safety tasks, our method significantly surpasses several competitive baselines, including random weighting, uniform weighting, single-head Bradley-Terry models, and LLM-based judging methods. Our proposed approach is training-free, broadly applicable to various datasets, and maintains interpretability, offering a practical and effective solution for multi-attribute reward modeling.

Related papers

• Suan: Rectifying Direct Preference Safety Alignment in Large Language Models [15.119045051735633]
  We introduce Suan, a novel preference optimization algorithm.<n>We obtain more interpretable and robust training dynamics.<n>In evaluations, Suan achieves superior safety alignment while fully preserving response utility.
  arXiv  Detail & Related papers  (2026-09-08T12:04:38Z)
• Too Correct to Learn: Reinforcement Learning on Saturated Reasoning Data [55.84428098924793]
  We propose Constrained Uniform Top-K Sampling (CUTS), a parameter-free decoding strategy enforcing structure-preserving exploration.<n>We integrate this into Mixed-CUTS, a training framework synergizing exploitative and exploratory rollouts to amplify intra-group advantage variance.<n> Notably, Mixed-CUTS improves Pass@1 accuracy on the challenging AIME25 benchmark by up to 15.1% over standard GRPO.
  arXiv  Detail & Related papers  (2026-04-20T16:43:28Z)
• Approximating Human Preferences Using a Multi-Judge Learned System [35.18016233072556]
  We propose a framework for modeling diverse, persona-based preferences by learning to aggregate outputs from multiple rubric-conditioned judges.<n>Our contributions include a persona-based method for synthesizing preference labels at scale and two distinct implementations of our aggregator.
  arXiv  Detail & Related papers  (2025-10-29T18:32:53Z)
• RiskPO: Risk-based Policy Optimization via Verifiable Reward for LLM Post-Training [13.309653291779233]
  Reinforcement learning with verifiable reward has emerged as a central paradigm for post-training large language models (LLMs)<n>We argue that these issues stem from overemphasizing high-probability output sequences while neglecting rare but informative reasoning paths.<n>We propose Risk-based Policy Optimization (RiskPO), which substitutes classical mean-based objectives with principled risk measures.
  arXiv  Detail & Related papers  (2025-10-01T13:53:09Z)
• IntentionReasoner: Facilitating Adaptive LLM Safeguards through Intent Reasoning and Selective Query Refinement [35.904652937034136]
  We introduce IntentionReasoner, a novel safeguard mechanism that leverages a dedicated guard model to perform intent reasoning.<n>We show that IntentionReasoner excels in multiple safeguard benchmarks, generation quality evaluations, and jailbreak attack scenarios.
  arXiv  Detail & Related papers  (2025-08-27T16:47:31Z)
• EDGE-GRPO: Entropy-Driven GRPO with Guided Error Correction for Advantage Diversity [7.818698554631196]
  Group Relative Policy Optimization (GRPO) algorithm relies on sparse reward rules, leading to the advantage collapse problem.<n>We propose the EDGE-GRPO algorithm, which adopts textbfEntropy-textbfDriven Advantage and textbfGuided textbfError Correction to effectively mitigate the problem of advantage collapse.
  arXiv  Detail & Related papers  (2025-07-29T14:23:58Z)
• Learning Pareto-Optimal Rewards from Noisy Preferences: A Framework for Multi-Objective Inverse Reinforcement Learning [0.0]
  We introduce a theoretical framework for preference-based Multi-Objective Inverse Reinforcement Learning (MO-IRL), where human preferences are modeled as latent vector-valued reward functions.<n>Our results bridge the gap between practical alignment techniques and theoretical guarantees, providing a principled foundation for learning aligned behaviors.
  arXiv  Detail & Related papers  (2025-05-17T06:09:13Z)
• Advancing Embodied Agent Security: From Safety Benchmarks to Input Moderation [52.83870601473094]
  Embodied agents exhibit immense potential across a multitude of domains. Existing research predominantly concentrates on the security of general large language models. This paper introduces a novel input moderation framework, meticulously designed to safeguard embodied agents.
  arXiv  Detail & Related papers  (2025-04-22T08:34:35Z)
• More is Less: The Pitfalls of Multi-Model Synthetic Preference Data in DPO Safety Alignment [80.04449725137177]
  Direct Preference Optimization (DPO) has emerged as a simple, yet effective alternative to reinforcement learning from human feedback. Our study reveals a striking, safety-specific phenomenon associated with DPO alignment. Using solely self-generated responses for both chosen and rejected pairs significantly outperforms configurations that incorporate responses from stronger models.
  arXiv  Detail & Related papers  (2025-04-03T00:36:40Z)
• Representation-based Reward Modeling for Efficient Safety Alignment of Large Language Model [84.00480999255628]
  Reinforcement Learning algorithms for safety alignment of Large Language Models (LLMs) encounter the challenge of distribution shift.<n>Current approaches typically address this issue through online sampling from the target policy.<n>We propose a new framework that leverages the model's intrinsic safety judgment capability to extract reward signals.
  arXiv  Detail & Related papers  (2025-03-13T06:40:34Z)
• Data-adaptive Safety Rules for Training Reward Models [9.172858508920696]
  Reinforcement Learning from Human Feedback (RLHF) is commonly employed to tailor models to human preferences.<n>In this paper, we propose a dynamic method that adaptively selects the most important rules for each response pair.<n>As of January 25, 2025, our model achieved the highest safety performance on the leaderboard, surpassing various larger models.
  arXiv  Detail & Related papers  (2025-01-26T08:49:46Z)
• REINFORCE++: An Efficient RLHF Algorithm with Robustness to Both Prompt and Reward Models [8.587685197004097]
  REINFORCE++ is a novel approach that removes the critic model while using the normalized reward of a batch as the baseline. It exhibits robust performance across various reward models without requiring prompt set truncation. It achieves superior generalization in both RLHF and long chain-of-thought settings compared to existing REINFORCE-based methods.
  arXiv  Detail & Related papers  (2025-01-04T02:08:06Z)
• Inverse-RLignment: Large Language Model Alignment from Demonstrations through Inverse Reinforcement Learning [62.05713042908654]
  We introduce Alignment from Demonstrations (AfD), a novel approach leveraging high-quality demonstration data to overcome these challenges.<n>We formalize AfD within a sequential decision-making framework, highlighting its unique challenge of missing reward signals.<n> Practically, we propose a computationally efficient algorithm that extrapolates over a tailored reward model for AfD.
  arXiv  Detail & Related papers  (2024-05-24T15:13:53Z)
• LIRE: listwise reward enhancement for preference alignment [27.50204023448716]
  We propose a gradient-based reward optimization approach that incorporates the offline rewards of multiple responses into a streamlined listwise framework. LIRE is straightforward to implement, requiring minimal parameter tuning, and seamlessly aligns with the pairwise paradigm. Our experiments demonstrate that LIRE consistently outperforms existing methods across several benchmarks on dialogue and summarization tasks.
  arXiv  Detail & Related papers  (2024-05-22T10:21:50Z)
• RLHF from Heterogeneous Feedback via Personalization and Preference Aggregation [24.374185140811115]
  Reinforcement learning from human feedback (RLHF) has been an effective technique for aligning AI systems with human values. In this paper, we focus on addressing the issues due to the inherent heterogeneity in human preferences, as well as their potential strategic behavior in providing feedback. We propose two frameworks to address heterogeneous human feedback in principled ways: personalization-based one and aggregation-based one.
  arXiv  Detail & Related papers  (2024-04-30T23:57:23Z)
• Improving Reinforcement Learning from Human Feedback Using Contrastive Rewards [26.40009657912622]
  Reinforcement learning from human feedback (RLHF) is the mainstream paradigm used to align large language models (LLMs) with human preferences. Yet existing RLHF heavily relies on accurate and informative reward models, which are vulnerable and sensitive to noise from various sources. In this work, we improve the effectiveness of the reward model by introducing a penalty term on the reward, named as textitcontrastive rewards
  arXiv  Detail & Related papers  (2024-03-12T14:51:57Z)
• Reinforcement Learning from Diverse Human Preferences [68.4294547285359]
  This paper develops a method for crowd-sourcing preference labels and learning from diverse human preferences. The proposed method is tested on a variety of tasks in DMcontrol and Meta-world. It has shown consistent and significant improvements over existing preference-based RL algorithms when learning from diverse feedback.
  arXiv  Detail & Related papers  (2023-01-27T15:18:54Z)
• Choosing the Best of Both Worlds: Diverse and Novel Recommendations through Multi-Objective Reinforcement Learning [68.45370492516531]
  We introduce Scalarized Multi-Objective Reinforcement Learning (SMORL) for the Recommender Systems (RS) setting. SMORL agent augments standard recommendation models with additional RL layers that enforce it to simultaneously satisfy three principal objectives: accuracy, diversity, and novelty of recommendations. Our experimental results on two real-world datasets reveal a substantial increase in aggregate diversity, a moderate increase in accuracy, reduced repetitiveness of recommendations, and demonstrate the importance of reinforcing diversity and novelty as complementary objectives.
  arXiv  Detail & Related papers  (2021-10-28T13:22:45Z)

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.