paper-with-me

홈 › Papers

REAL: Response Embedding-based Alignment for LLMs

2024-09-17 · Honggen Zhang, Xufeng Zhao, Igor Molybog, June Zhang

Aligning large language models (LLMs) to human preferences is a crucial step in building helpful and safe AI tools, which usually involve training on supervised datasets. Popular algorithms such as Direct Preference Optimization rely on pairs of AI-generated responses ranked according to human feedback. The response pair annotation process is the most labor-intensive and costly part of the alignment pipeline, and improving its efficiency and annotation quality would have a meaningful impact on AI development. We propose REAL: Response Embedding-based Alignment for LLMs, a strategy for constructing a high-quality training dataset that focuses on acquiring the most informative response pairs for labeling out of a set of response candidates. Our selection process is based on embedding responses independently of prompts. Experimental results on real-world dataset SHP2 and synthetic HH-RLHF benchmarks indicate that choosing dissimilar response pairs enhances the direct alignment of LLMs while reducing inherited labeling errors. The model aligned on dissimilar response pairs obtained a better margin and win rate on the dialogue task. Our findings suggest that focusing on distinct pairs can reduce the label error to improve the efficiency of LLM alignment, saving up to 65% of annotators' work.

📄 PDF Abstract BibTeX arXiv:2409.17169

Code (1)

honggen-zhang/real-alignment 공식 구현 pytorch

Methods 이 논문이 사용한 방법론

SET Dynamic Sparse Training method where weight mask is updated randomly periodically

Similar Papers 제목 키워드 기반

Correlating instruction-tuning (in multimodal models) with vision-language processing (in the brain)

2025-05-26 · Subba Reddy Oota, Akshett Jindal, Ishani Mondal, Khushbu Pahwa 외

Transformer-based language models, though not explicitly trained to mimic brain recordings, have demonstrated surprising alignment with brain activity. Progress in these models-through increased size, instruction-tuning,…

Image Captioning

Following the Autoregressive Nature of LLM Embeddings via Compression and Alignment

2025-02-17 · Jingcheng Deng, Zhongtao Jiang, Liang Pang, Liwei Chen 외

A new trend uses LLMs as dense text encoders via contrastive learning. However, since LLM embeddings predict the probability distribution of the next token, they are inherently generative and distributive, conflicting wi…

Contrastive Learning

DROJ: A Prompt-Driven Attack against Large Language Models

2024-11-14 · Leyang Hu, Boran Wang

Large Language Models (LLMs) have demonstrated exceptional capabilities across various natural language processing tasks. Due to their training on internet-sourced datasets, LLMs can sometimes generate objectionable cont…

Embedding Poisoning: Bypassing Safety Alignment via Embedding Semantic Shift

2025-09-08 · Shuai Yuan, Zhibo Zhang, Yuxi Li, Guangdong Bai 외 arxiv

The widespread distribution of Large Language Models (LLMs) through public platforms like Hugging Face introduces significant security challenges. While these platforms perform basic security scans, they often fail to de…

Are LLMs (Really) Ideological? An IRT-based Analysis and Alignment Tool for Perceived Socio-Economic Bias in LLMs

2025-03-17 · Jasmin Wachter, Michael Radloff, Maja Smolej, Katharina Kinder-Kurlanda

We introduce an Item Response Theory (IRT)-based framework to detect and quantify socioeconomic bias in large language models (LLMs) without relying on subjective human judgments. Unlike traditional methods, IRT accounts…