Salvato in:
Dettagli Bibliografici
Autori principali: Wang, Ziqi, Zhang, Hanlin, Li, Xiner, Huang, Kuan-Hao, Han, Chi, Ji, Shuiwang, Kakade, Sham M., Peng, Hao, Ji, Heng
Natura: Preprint
Pubblicazione: 2024
Soggetti:
Accesso online:https://arxiv.org/abs/2407.01100
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866910900020051968
author Wang, Ziqi
Zhang, Hanlin
Li, Xiner
Huang, Kuan-Hao
Han, Chi
Ji, Shuiwang
Kakade, Sham M.
Peng, Hao
Ji, Heng
author_facet Wang, Ziqi
Zhang, Hanlin
Li, Xiner
Huang, Kuan-Hao
Han, Chi
Ji, Shuiwang
Kakade, Sham M.
Peng, Hao
Ji, Heng
contents Position bias has proven to be a prevalent issue of modern language models (LMs), where the models prioritize content based on its position within the given context. This bias often leads to unexpected model failures and hurts performance, robustness, and reliability across various applications. Our mechanistic analysis attributes the position bias to two components employed in nearly all state-of-the-art LMs: causal attention and relative positional encodings. Based on the analyses, we propose to eliminate position bias (e.g., different retrieved documents' orders in QA affect performance) with a training-free zero-shot approach. Our method changes the causal attention to bidirectional attention between documents and utilizes model attention values to decide the relative orders of documents instead of using the order provided in input prompts, therefore enabling Position-INvariant inferencE (PINE) at the document level. By eliminating position bias, models achieve better performance and reliability in downstream tasks, including LM-as-a-judge, retrieval-augmented QA, molecule generation, and math reasoning. Notably, PINE is especially useful when adapting LMs for evaluating reasoning pairs: it consistently provides 8 to 10 percentage points performance gains, making Llama-3-70B-Instruct perform even better than GPT-4-0125-preview and GPT-4o-2024-08-06 on the RewardBench reasoning set.
format Preprint
id arxiv_https___arxiv_org_abs_2407_01100
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Eliminating Position Bias of Language Models: A Mechanistic Approach
Wang, Ziqi
Zhang, Hanlin
Li, Xiner
Huang, Kuan-Hao
Han, Chi
Ji, Shuiwang
Kakade, Sham M.
Peng, Hao
Ji, Heng
Computation and Language
Machine Learning
Position bias has proven to be a prevalent issue of modern language models (LMs), where the models prioritize content based on its position within the given context. This bias often leads to unexpected model failures and hurts performance, robustness, and reliability across various applications. Our mechanistic analysis attributes the position bias to two components employed in nearly all state-of-the-art LMs: causal attention and relative positional encodings. Based on the analyses, we propose to eliminate position bias (e.g., different retrieved documents' orders in QA affect performance) with a training-free zero-shot approach. Our method changes the causal attention to bidirectional attention between documents and utilizes model attention values to decide the relative orders of documents instead of using the order provided in input prompts, therefore enabling Position-INvariant inferencE (PINE) at the document level. By eliminating position bias, models achieve better performance and reliability in downstream tasks, including LM-as-a-judge, retrieval-augmented QA, molecule generation, and math reasoning. Notably, PINE is especially useful when adapting LMs for evaluating reasoning pairs: it consistently provides 8 to 10 percentage points performance gains, making Llama-3-70B-Instruct perform even better than GPT-4-0125-preview and GPT-4o-2024-08-06 on the RewardBench reasoning set.
title Eliminating Position Bias of Language Models: A Mechanistic Approach
topic Computation and Language
Machine Learning
url https://arxiv.org/abs/2407.01100