Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

DoLa: Decoding by Contrasting Layers Improves Factuality in Large Language Models

About

Despite their impressive capabilities, large language models (LLMs) are prone to hallucinations, i.e., generating content that deviates from facts seen during pretraining. We propose a simple decoding strategy for reducing hallucinations with pretrained LLMs that does not require conditioning on retrieved external knowledge nor additional fine-tuning. Our approach obtains the next-token distribution by contrasting the differences in logits obtained from projecting the later layers versus earlier layers to the vocabulary space, exploiting the fact that factual knowledge in an LLMs has generally been shown to be localized to particular transformer layers. We find that this Decoding by Contrasting Layers (DoLa) approach is able to better surface factual knowledge and reduce the generation of incorrect facts. DoLa consistently improves the truthfulness across multiple choices tasks and open-ended generation tasks, for example improving the performance of LLaMA family models on TruthfulQA by 12-17% absolute points, demonstrating its potential in making LLMs reliably generate truthful facts.

Yung-Sung Chuang, Yujia Xie, Hongyin Luo, Yoon Kim, James Glass, Pengcheng He• 2023

Related benchmarks

TaskDatasetResultRank
Object Hallucination EvaluationPOPE
Accuracy83.1
2019
Visual Question AnsweringVizWiz
Accuracy74.38
1820
Visual Question AnsweringTextVQA
Accuracy56.6
1453
Visual Question AnsweringVQA v2
Accuracy80.75
1429
Visual Question AnsweringGQA
Accuracy73.4
1425
Mathematical ReasoningGSM8K
Accuracy90.75
1398
Code GenerationHumanEval
Pass@112.8
1043
Text-based Visual Question AnsweringTextVQA
Accuracy84.19
962
Multimodal UnderstandingMMBench
Accuracy79.03
847
Instruction FollowingIFEval--
836
Showing 10 of 241 rows
...

Other info

Follow for update