CDR: Customizable Density Ratios of Strong-over-weak LLMs for Preference Annotation

• URL: http://arxiv.org/abs/2411.02481v2
• Date: Mon, 11 Nov 2024 17:34:00 GMT
• Title: CDR: Customizable Density Ratios of Strong-over-weak LLMs for Preference Annotation
• Authors: Guangxuan Xu, Kai Xu, Shivchander Sudalairaj, Hao Wang, Akash Srivastava,
• Abstract summary: Preference tuning of large language models (LLMs) relies on high-quality human preference data. We introduce customized density ratio (CDR), a training-free and highly effective method that leverages off-the-shelf LLMs for preference data annotation. We show that tailoring the density ratio reward function with specific criteria and preference exemplars enhances performance across domains and within target areas.
• Score: 15.776175440446414
• License:
• Abstract: Preference tuning of large language models (LLMs) relies on high-quality human preference data, which is often expensive and time-consuming to gather. While existing methods can use trained reward models or proprietary model as judges for preference annotation, they have notable drawbacks: training reward models remain dependent on initial human data, and using proprietary model imposes license restrictions that inhibits commercial usage. In this paper, we introduce customized density ratio (CDR), a training-free and highly effective method that leverages off-the-shelf LLMs for preference data annotation. Our approach uses the log-density ratio between a better-aligned LLM and a less aligned LLM as a reward signal. We explores 221 different LLMs pairs and empirically demonstrate that increasing the performance gap between paired LLMs correlates with better reward generalization. Furthermore, we show that tailoring the density ratio reward function with specific criteria and preference exemplars enhances performance across domains and within target areas. In our experiment using density ratio from a pair of Mistral-7B models, CDR achieves a RewardBench score of 82.6, outperforming the best trained reward functions from same model class and demonstrating competitive performance against SoTA models in Safety (91.0) and Reasoning (88.0) domains. We use CDR to annotate an on-policy preference dataset with which we preference tune Llama-3-8B-Instruct with SimPO. Using reward signals from two relatively weak models, our approach pushes Llama-3-8B to achieve a 37.4% (+15.1%) win rate on ArenaHard and a 40.7% (+17.8%) win rate on Length-Controlled AlpacaEval 2.0, along with a score of 8.0 on MT-Bench.

Related papers

• Margin Matching Preference Optimization: Enhanced Model Alignment with Granular Feedback [64.67540769692074]
  Large language models (LLMs) fine-tuned with alignment techniques, such as reinforcement learning from human feedback, have been instrumental in developing some of the most capable AI systems to date. We introduce an approach called Margin Matching Preference Optimization (MMPO), which incorporates relative quality margins into optimization, leading to improved LLM policies and reward models. Experiments with both human and AI feedback data demonstrate that MMPO consistently outperforms baseline methods, often by a substantial margin, on popular benchmarks including MT-bench and RewardBench.
  arXiv  Detail & Related papers  (2024-10-04T04:56:11Z)
• Generative Reward Models [42.30530024761532]
  Reinforcement Learning from Human Feedback (RLHF) has greatly improved the performance of modern Large Language Models (LLMs) Recent work has shown that synthetic preferences labels may not align well with human preference judgments. We propose a hybrid approach that unifies RLHF and RLAIF methodologies. Our results show that combining the strengths of RLHF and RLAIF offers a promising approach for improving the quality of synthetic preference labels.
  arXiv  Detail & Related papers  (2024-10-02T17:58:39Z)
• RRM: Robust Reward Model Training Mitigates Reward Hacking [51.12341734942797]
  Reward models (RMs) play a pivotal role in aligning large language models with human preferences. We introduce a causal framework that learns preferences independent of these artifacts. Experiments show that our approach successfully filters out undesirable artifacts, yielding a more robust reward model.
  arXiv  Detail & Related papers  (2024-09-20T01:46:07Z)
• Constructing Domain-Specific Evaluation Sets for LLM-as-a-judge [15.980606104936365]
  Large Language Models (LLMs) have revolutionized the landscape of machine learning, yet current benchmarks often fall short in capturing the diverse behavior of these models in real-world applications. Existing frameworks like Alpaca-Eval 2.0 LC citedubois2024lengthcontrolledalpacaevalsimpleway and Arena-Hard v0.1 citeli2024crowdsourced are limited by their focus on general-purpose queries and lack of diversity across domains such as law, medicine, and multilingual contexts. We introduce a novel data pipeline that curates, domain-specific evaluation sets tailored for LLM-as
  arXiv  Detail & Related papers  (2024-08-16T15:41:43Z)
• Anchored Preference Optimization and Contrastive Revisions: Addressing Underspecification in Alignment [57.03947082589616]
  Large Language Models (LLMs) are often aligned using contrastive alignment objectives and preference pair datasets. We study this and find that preference data gives a better learning signal when the underlying responses are contrastive. We introduce Contrastive Learning from AI Revisions (CLAIR), a data-creation method which leads to more contrastive preference pairs. Our best model, trained on 32K CLAIR preferences with APO, improves Llama-3-8B-Instruct by 7.65%, closing the gap with GPT4-turbo by 45%.
  arXiv  Detail & Related papers  (2024-08-12T16:24:51Z)
• Closing the gap between open-source and commercial large language models for medical evidence summarization [20.60798771155072]
  Large language models (LLMs) hold great promise in summarizing medical evidence. Most recent studies focus on the application of proprietary LLMs. While open-source LLMs allow better transparency and customization, their performance falls short compared to proprietary ones.
  arXiv  Detail & Related papers  (2024-07-25T05:03:01Z)
• Foundational Autoraters: Taming Large Language Models for Better Automatic Evaluation [20.41379322900742]
  We introduce FLAMe, a family of Foundational Large Autorater Models. FLAMe is trained on our large and diverse collection of 100+ quality assessment tasks. We show that FLAMe can also serve as a powerful starting point for further downstream fine-tuning.
  arXiv  Detail & Related papers  (2024-07-15T15:33:45Z)
• Monte Carlo Tree Search Boosts Reasoning via Iterative Preference Learning [55.96599486604344]
  We introduce an approach aimed at enhancing the reasoning capabilities of Large Language Models (LLMs) through an iterative preference learning process. We use Monte Carlo Tree Search (MCTS) to iteratively collect preference data, utilizing its look-ahead ability to break down instance-level rewards into more granular step-level signals. The proposed algorithm employs Direct Preference Optimization (DPO) to update the LLM policy using this newly generated step-level preference data.
  arXiv  Detail & Related papers  (2024-05-01T11:10:24Z)
• Dissecting Human and LLM Preferences [80.55271307662365]
  We find that humans are less sensitive to errors, favor responses that support their stances, and show clear dislike when models admit their limits. advanced LLMs like GPT-4-Turbo emphasize correctness, clarity, and harmlessness more. We show that preference-based evaluation can be intentionally manipulated.
  arXiv  Detail & Related papers  (2024-02-17T14:34:31Z)
• WARM: On the Benefits of Weight Averaged Reward Models [63.08179139233774]
  We propose Weight Averaged Reward Models (WARM) to mitigate reward hacking. Experiments on summarization tasks, using best-of-N and RL methods, shows that WARM improves the overall quality and alignment of LLM predictions.
  arXiv  Detail & Related papers  (2024-01-22T18:27:08Z)

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.