paper-with-me

홈 › Papers

Scalable Ensembling For Mitigating Reward Overoptimisation

2024-06-03 · Ahmed M. Ahmed, Rafael Rafailov, Stepan Sharkov, Xuechen Li, Sanmi Koyejo

Reinforcement Learning from Human Feedback (RLHF) has enabled significant advancements within language modeling for powerful, instruction-following models. However, the alignment of these models remains a pressing challenge as the policy tends to overfit the learned `proxy" reward model past an inflection point of utility as measured by a `gold" reward model that is more performant -- a phenomenon known as overoptimisation. Prior work has mitigated this issue by computing a pessimistic statistic over an ensemble of reward models, which is common in Offline Reinforcement Learning but incredibly costly for language models with high memory requirements, making such approaches infeasible for sufficiently large models. To this end, we propose using a shared encoder but separate linear heads. We find this leads to similar performance as the full ensemble while allowing tremendous savings in memory and time required for training for models of similar size.

📄 PDF Abstract BibTeX arXiv:2406.01013

Code (0)

등록된 구현이 없습니다.

Tasks

Instruction FollowingLanguage ModelingLanguage Modellingreinforcement-learningReinforcement Learning

Similar Papers 제목 키워드 기반

SCE: Scalable Consistency Ensembles Make Blackbox Large Language Model Generation More Reliable

2025-03-13 · Jiaxin Zhang, Zhuohang Li, Wendi Cui, Kamalika Das 외

Large language models (LLMs) have demonstrated remarkable performance, yet their diverse strengths and weaknesses prevent any single LLM from achieving dominance across all tasks. Ensembling multiple LLMs is a promising …

Language ModelingLanguage ModellingLarge Language Model

Merging Multi-Task Models via Weight-Ensembling Mixture of Experts

2024-02-01 · Anke Tang, Li Shen, Yong Luo, Nan Yin 외

Merging various task-specific Transformer-based models trained on different tasks into a single unified model can execute all the tasks concurrently. Previous methods, exemplified by task arithmetic, have been proven to …

Mixture-of-ExpertsTask Arithmetic

On Cost-Effective LLM-as-a-Judge Improvement Techniques

2026-04-15 · Ryan Lail, Luke Markham arxiv

Using a language model to score or rank candidate responses has become a scalable alternative to human evaluation in reinforcement learning from human feedback (RLHF) pipelines, benchmarking, and application layer evalua…

Reinforcement Learning

Purifying Large Language Models by Ensembling a Small Language Model

2024-02-19 · Tianlin Li, Qian Liu, Tianyu Pang, Chao Du 외

The emerging success of large language models (LLMs) heavily relies on collecting abundant training data from external (untrusted) sources. Despite substantial efforts devoted to data cleaning and curation, well-construc…

Data PoisoningLanguage ModelingLanguage ModellingSmall Language Model

ARES: A Reading Comprehension Ensembling Service

2020-10-01 · EMNLP 2020 11 · Anthony Ferritto, Lin Pan, Rishav Chakravarti, Salim Roukos 외

We introduce ARES (A Reading Comprehension Ensembling Service): a novel Machine Reading Comprehension (MRC) demonstration system which utilizes an ensemble of models to increase F1 by 2.3 points. While many of the top le…

Machine Reading ComprehensionNatural QuestionsQuestion AnsweringReading Comprehension