Our new X account is live! Follow @wizwand_team for updates
WorkDL logo mark

Discerning and Resolving Knowledge Conflicts through Adaptive Decoding with Contextual Information-Entropy Constraint

About

Large language models internalize enormous parametric knowledge during pre-training. Concurrently, realistic applications necessitate external contextual knowledge to aid models on the underlying tasks. This raises a crucial dilemma known as knowledge conflicts, where the contextual knowledge clashes with the However, existing decoding works are specialized in resolving knowledge conflicts and could inadvertently deteriorate performance in absence of conflicts. In this paper, we propose an adaptive decoding method, termed as contextual information-entropy constraint decoding (COIECD), to discern whether the knowledge conflicts occur and resolve them. It can improve the model's faithfulness to conflicting context, and simultaneously maintain high performance among non- Our experiments show that COIECD exhibits strong performance and robustness over knowledge conflicts in realistic datasets. Code is available.

Xiaowei Yuan, Zhao Yang, Yequan Wang, Shengping Liu, Jun Zhao, Kang Liu• 2024

Related benchmarks

TaskDatasetResultRank
Abstractive Text SummarizationCNN/Daily Mail (test)
ROUGE-L21.17
169
Question AnsweringSQuAD
F184.99
127
Question AnsweringTriviaQA
EM80.95
116
Question AnsweringNQ
EM48.84
57
Question AnsweringSQuAD
Exact Match88.9
50
Abstractive SummarizationXSum (test)
ROUGE-L15.77
44
Question AnsweringMuSiQue
Accuracy (ACC)70.5
36
Open-book generation under knowledge conflictConFiQA 1,500 subset
Ps Score71
32
Question AnsweringFaithEval
Accuracy67.7
27
Question AnsweringRealtimeQA
Accuracy78.8
27
Showing 10 of 23 rows

Other info

Code

Follow for update